Searched +hist:3 +hist:b32123d (Results 1 - 5 of 5) sorted by path
/linux-master/mm/ | ||
H A D | hugetlb.c | diff 52ccdde1 Fri Apr 19 02:58:19 MDT 2024 Miaohe Lin <linmiaohe@huawei.com> mm/hugetlb: fix DEBUG_LOCKS_WARN_ON(1) when dissolve_free_hugetlb_folio() When I did memory failure tests recently, below warning occurs: DEBUG_LOCKS_WARN_ON(1) WARNING: CPU: 8 PID: 1011 at kernel/locking/lockdep.c:232 __lock_acquire+0xccb/0x1ca0 Modules linked in: mce_inject hwpoison_inject CPU: 8 PID: 1011 Comm: bash Kdump: loaded Not tainted 6.9.0-rc3-next-20240410-00012-gdb69f219f4be #3 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014 RIP: 0010:__lock_acquire+0xccb/0x1ca0 RSP: 0018:ffffa7a1c7fe3bd0 EFLAGS: 00000082 RAX: 0000000000000000 RBX: eb851eb853975fcf RCX: ffffa1ce5fc1c9c8 RDX: 00000000ffffffd8 RSI: 0000000000000027 RDI: ffffa1ce5fc1c9c0 RBP: ffffa1c6865d3280 R08: ffffffffb0f570a8 R09: 0000000000009ffb R10: 0000000000000286 R11: ffffffffb0f2ad50 R12: ffffa1c6865d3d10 R13: ffffa1c6865d3c70 R14: 0000000000000000 R15: 0000000000000004 FS: 00007ff9f32aa740(0000) GS:ffffa1ce5fc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007ff9f3134ba0 CR3: 00000008484e4000 CR4: 00000000000006f0 Call Trace: <TASK> lock_acquire+0xbe/0x2d0 _raw_spin_lock_irqsave+0x3a/0x60 hugepage_subpool_put_pages.part.0+0xe/0xc0 free_huge_folio+0x253/0x3f0 dissolve_free_huge_page+0x147/0x210 __page_handle_poison+0x9/0x70 memory_failure+0x4e6/0x8c0 hard_offline_page_store+0x55/0xa0 kernfs_fop_write_iter+0x12c/0x1d0 vfs_write+0x380/0x540 ksys_write+0x64/0xe0 do_syscall_64+0xbc/0x1d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7ff9f3114887 RSP: 002b:00007ffecbacb458 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007ff9f3114887 RDX: 000000000000000c RSI: 0000564494164e10 RDI: 0000000000000001 RBP: 0000564494164e10 R08: 00007ff9f31d1460 R09: 000000007fffffff R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c R13: 00007ff9f321b780 R14: 00007ff9f3217600 R15: 00007ff9f3216a00 </TASK> Kernel panic - not syncing: kernel: panic_on_warn set ... CPU: 8 PID: 1011 Comm: bash Kdump: loaded Not tainted 6.9.0-rc3-next-20240410-00012-gdb69f219f4be #3 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> panic+0x326/0x350 check_panic_on_warn+0x4f/0x50 __warn+0x98/0x190 report_bug+0x18e/0x1a0 handle_bug+0x3d/0x70 exc_invalid_op+0x18/0x70 asm_exc_invalid_op+0x1a/0x20 RIP: 0010:__lock_acquire+0xccb/0x1ca0 RSP: 0018:ffffa7a1c7fe3bd0 EFLAGS: 00000082 RAX: 0000000000000000 RBX: eb851eb853975fcf RCX: ffffa1ce5fc1c9c8 RDX: 00000000ffffffd8 RSI: 0000000000000027 RDI: ffffa1ce5fc1c9c0 RBP: ffffa1c6865d3280 R08: ffffffffb0f570a8 R09: 0000000000009ffb R10: 0000000000000286 R11: ffffffffb0f2ad50 R12: ffffa1c6865d3d10 R13: ffffa1c6865d3c70 R14: 0000000000000000 R15: 0000000000000004 lock_acquire+0xbe/0x2d0 _raw_spin_lock_irqsave+0x3a/0x60 hugepage_subpool_put_pages.part.0+0xe/0xc0 free_huge_folio+0x253/0x3f0 dissolve_free_huge_page+0x147/0x210 __page_handle_poison+0x9/0x70 memory_failure+0x4e6/0x8c0 hard_offline_page_store+0x55/0xa0 kernfs_fop_write_iter+0x12c/0x1d0 vfs_write+0x380/0x540 ksys_write+0x64/0xe0 do_syscall_64+0xbc/0x1d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7ff9f3114887 RSP: 002b:00007ffecbacb458 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007ff9f3114887 RDX: 000000000000000c RSI: 0000564494164e10 RDI: 0000000000000001 RBP: 0000564494164e10 R08: 00007ff9f31d1460 R09: 000000007fffffff R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c R13: 00007ff9f321b780 R14: 00007ff9f3217600 R15: 00007ff9f3216a00 </TASK> After git bisecting and digging into the code, I believe the root cause is that _deferred_list field of folio is unioned with _hugetlb_subpool field. In __update_and_free_hugetlb_folio(), folio->_deferred_list is initialized leading to corrupted folio->_hugetlb_subpool when folio is hugetlb. Later free_huge_folio() will use _hugetlb_subpool and above warning happens. But it is assumed hugetlb flag must have been cleared when calling folio_put() in update_and_free_hugetlb_folio(). This assumption is broken due to below race: CPU1 CPU2 dissolve_free_huge_page update_and_free_pages_bulk update_and_free_hugetlb_folio hugetlb_vmemmap_restore_folios folio_clear_hugetlb_vmemmap_optimized clear_flag = folio_test_hugetlb_vmemmap_optimized if (clear_flag) <-- False, it's already cleared. __folio_clear_hugetlb(folio) <-- Hugetlb is not cleared. folio_put free_huge_folio <-- free_the_page is expected. list_for_each_entry() __folio_clear_hugetlb <-- Too late. Fix this issue by checking whether folio is hugetlb directly instead of checking clear_flag to close the race window. Link: https://lkml.kernel.org/r/20240419085819.1901645-1-linmiaohe@huawei.com Fixes: 32c877191e02 ("hugetlb: do not clear hugetlb dtor until allocating vmemmap") Signed-off-by: Miaohe Lin <linmiaohe@huawei.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 52ccdde1 Fri Apr 19 02:58:19 MDT 2024 Miaohe Lin <linmiaohe@huawei.com> mm/hugetlb: fix DEBUG_LOCKS_WARN_ON(1) when dissolve_free_hugetlb_folio() When I did memory failure tests recently, below warning occurs: DEBUG_LOCKS_WARN_ON(1) WARNING: CPU: 8 PID: 1011 at kernel/locking/lockdep.c:232 __lock_acquire+0xccb/0x1ca0 Modules linked in: mce_inject hwpoison_inject CPU: 8 PID: 1011 Comm: bash Kdump: loaded Not tainted 6.9.0-rc3-next-20240410-00012-gdb69f219f4be #3 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014 RIP: 0010:__lock_acquire+0xccb/0x1ca0 RSP: 0018:ffffa7a1c7fe3bd0 EFLAGS: 00000082 RAX: 0000000000000000 RBX: eb851eb853975fcf RCX: ffffa1ce5fc1c9c8 RDX: 00000000ffffffd8 RSI: 0000000000000027 RDI: ffffa1ce5fc1c9c0 RBP: ffffa1c6865d3280 R08: ffffffffb0f570a8 R09: 0000000000009ffb R10: 0000000000000286 R11: ffffffffb0f2ad50 R12: ffffa1c6865d3d10 R13: ffffa1c6865d3c70 R14: 0000000000000000 R15: 0000000000000004 FS: 00007ff9f32aa740(0000) GS:ffffa1ce5fc00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007ff9f3134ba0 CR3: 00000008484e4000 CR4: 00000000000006f0 Call Trace: <TASK> lock_acquire+0xbe/0x2d0 _raw_spin_lock_irqsave+0x3a/0x60 hugepage_subpool_put_pages.part.0+0xe/0xc0 free_huge_folio+0x253/0x3f0 dissolve_free_huge_page+0x147/0x210 __page_handle_poison+0x9/0x70 memory_failure+0x4e6/0x8c0 hard_offline_page_store+0x55/0xa0 kernfs_fop_write_iter+0x12c/0x1d0 vfs_write+0x380/0x540 ksys_write+0x64/0xe0 do_syscall_64+0xbc/0x1d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7ff9f3114887 RSP: 002b:00007ffecbacb458 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007ff9f3114887 RDX: 000000000000000c RSI: 0000564494164e10 RDI: 0000000000000001 RBP: 0000564494164e10 R08: 00007ff9f31d1460 R09: 000000007fffffff R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c R13: 00007ff9f321b780 R14: 00007ff9f3217600 R15: 00007ff9f3216a00 </TASK> Kernel panic - not syncing: kernel: panic_on_warn set ... CPU: 8 PID: 1011 Comm: bash Kdump: loaded Not tainted 6.9.0-rc3-next-20240410-00012-gdb69f219f4be #3 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> panic+0x326/0x350 check_panic_on_warn+0x4f/0x50 __warn+0x98/0x190 report_bug+0x18e/0x1a0 handle_bug+0x3d/0x70 exc_invalid_op+0x18/0x70 asm_exc_invalid_op+0x1a/0x20 RIP: 0010:__lock_acquire+0xccb/0x1ca0 RSP: 0018:ffffa7a1c7fe3bd0 EFLAGS: 00000082 RAX: 0000000000000000 RBX: eb851eb853975fcf RCX: ffffa1ce5fc1c9c8 RDX: 00000000ffffffd8 RSI: 0000000000000027 RDI: ffffa1ce5fc1c9c0 RBP: ffffa1c6865d3280 R08: ffffffffb0f570a8 R09: 0000000000009ffb R10: 0000000000000286 R11: ffffffffb0f2ad50 R12: ffffa1c6865d3d10 R13: ffffa1c6865d3c70 R14: 0000000000000000 R15: 0000000000000004 lock_acquire+0xbe/0x2d0 _raw_spin_lock_irqsave+0x3a/0x60 hugepage_subpool_put_pages.part.0+0xe/0xc0 free_huge_folio+0x253/0x3f0 dissolve_free_huge_page+0x147/0x210 __page_handle_poison+0x9/0x70 memory_failure+0x4e6/0x8c0 hard_offline_page_store+0x55/0xa0 kernfs_fop_write_iter+0x12c/0x1d0 vfs_write+0x380/0x540 ksys_write+0x64/0xe0 do_syscall_64+0xbc/0x1d0 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7ff9f3114887 RSP: 002b:00007ffecbacb458 EFLAGS: 00000246 ORIG_RAX: 0000000000000001 RAX: ffffffffffffffda RBX: 000000000000000c RCX: 00007ff9f3114887 RDX: 000000000000000c RSI: 0000564494164e10 RDI: 0000000000000001 RBP: 0000564494164e10 R08: 00007ff9f31d1460 R09: 000000007fffffff R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000c R13: 00007ff9f321b780 R14: 00007ff9f3217600 R15: 00007ff9f3216a00 </TASK> After git bisecting and digging into the code, I believe the root cause is that _deferred_list field of folio is unioned with _hugetlb_subpool field. In __update_and_free_hugetlb_folio(), folio->_deferred_list is initialized leading to corrupted folio->_hugetlb_subpool when folio is hugetlb. Later free_huge_folio() will use _hugetlb_subpool and above warning happens. But it is assumed hugetlb flag must have been cleared when calling folio_put() in update_and_free_hugetlb_folio(). This assumption is broken due to below race: CPU1 CPU2 dissolve_free_huge_page update_and_free_pages_bulk update_and_free_hugetlb_folio hugetlb_vmemmap_restore_folios folio_clear_hugetlb_vmemmap_optimized clear_flag = folio_test_hugetlb_vmemmap_optimized if (clear_flag) <-- False, it's already cleared. __folio_clear_hugetlb(folio) <-- Hugetlb is not cleared. folio_put free_huge_folio <-- free_the_page is expected. list_for_each_entry() __folio_clear_hugetlb <-- Too late. Fix this issue by checking whether folio is hugetlb directly instead of checking clear_flag to close the race window. Link: https://lkml.kernel.org/r/20240419085819.1901645-1-linmiaohe@huawei.com Fixes: 32c877191e02 ("hugetlb: do not clear hugetlb dtor until allocating vmemmap") Signed-off-by: Miaohe Lin <linmiaohe@huawei.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff b76b4690 Wed Apr 17 15:18:35 MDT 2024 Peter Xu <peterx@redhat.com> mm/hugetlb: fix missing hugetlb_lock for resv uncharge There is a recent report on UFFDIO_COPY over hugetlb: https://lore.kernel.org/all/000000000000ee06de0616177560@google.com/ 350: lockdep_assert_held(&hugetlb_lock); Should be an issue in hugetlb but triggered in an userfault context, where it goes into the unlikely path where two threads modifying the resv map together. Mike has a fix in that path for resv uncharge but it looks like the locking criteria was overlooked: hugetlb_cgroup_uncharge_folio_rsvd() will update the cgroup pointer, so it requires to be called with the lock held. Link: https://lkml.kernel.org/r/20240417211836.2742593-3-peterx@redhat.com Fixes: 79aa925bf239 ("hugetlb_cgroup: fix reservation accounting") Signed-off-by: Peter Xu <peterx@redhat.com> Reported-by: syzbot+4b8077a5fccc61c385a1@syzkaller.appspotmail.com Reviewed-by: Mina Almasry <almasrymina@google.com> Cc: David Hildenbrand <david@redhat.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff d5c3eb3f Thu Feb 22 07:04:15 MST 2024 Gang Li <gang.li@linux.dev> hugetlb: split hugetlb_hstate_alloc_pages 1G and 2M huge pages have different allocation and initialization logic, which leads to subtle differences in parallelization. Therefore, it is appropriate to split hugetlb_hstate_alloc_pages into gigantic and non-gigantic. This patch has no functional changes. Link: https://lkml.kernel.org/r/20240222140422.393911-3-gang.li@linux.dev Signed-off-by: Gang Li <ligang.bdlg@bytedance.com> Tested-by: David Rientjes <rientjes@google.com> Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com> Reviewed-by: Muchun Song <muchun.song@linux.dev> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Daniel Jordan <daniel.m.jordan@oracle.com> Cc: David Hildenbrand <david@redhat.com> Cc: Jane Chu <jane.chu@oracle.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Steffen Klassert <steffen.klassert@secunet.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff fc37bbb3 Thu Feb 22 07:04:14 MST 2024 Gang Li <gang.li@linux.dev> hugetlb: code clean for hugetlb_hstate_alloc_pages Patch series "hugetlb: parallelize hugetlb page init on boot", v6. Introduction ------------ Hugetlb initialization during boot takes up a considerable amount of time. For instance, on a 2TB system, initializing 1,800 1GB huge pages takes 1-2 seconds out of 10 seconds. Initializing 11,776 1GB pages on a 12TB Intel host takes more than 1 minute[1]. This is a noteworthy figure. Inspired by [2] and [3], hugetlb initialization can also be accelerated through parallelization. Kernel already has infrastructure like padata_do_multithreaded, this patch uses it to achieve effective results by minimal modifications. [1] https://lore.kernel.org/all/783f8bac-55b8-5b95-eb6a-11a583675000@google.com/ [2] https://lore.kernel.org/all/20200527173608.2885243-1-daniel.m.jordan@oracle.com/ [3] https://lore.kernel.org/all/20230906112605.2286994-1-usama.arif@bytedance.com/ [4] https://lore.kernel.org/all/76becfc1-e609-e3e8-2966-4053143170b6@google.com/ max_threads ----------- This patch use `padata_do_multithreaded` like this: ``` job.max_threads = num_node_state(N_MEMORY) * multiplier; padata_do_multithreaded(&job); ``` To fully utilize the CPU, the number of parallel threads needs to be carefully considered. `max_threads = num_node_state(N_MEMORY)` does not fully utilize the CPU, so we need to multiply it by a multiplier. Tests below indicate that a multiplier of 2 significantly improves performance, and although larger values also provide improvements, the gains are marginal. multiplier 1 2 3 4 5 ------------ ------- ------- ------- ------- ------- 256G 2node 358ms 215ms 157ms 134ms 126ms 2T 4node 979ms 679ms 543ms 489ms 481ms 50G 2node 71ms 44ms 37ms 30ms 31ms Therefore, choosing 2 as the multiplier strikes a good balance between enhancing parallel processing capabilities and maintaining efficient resource management. Test result ----------- test case no patch(ms) patched(ms) saved ------------------- -------------- ------------- -------- 256c2T(4 node) 1G 4745 2024 57.34% 128c1T(2 node) 1G 3358 1712 49.02% 12T 1G 77000 18300 76.23% 256c2T(4 node) 2M 3336 1051 68.52% 128c1T(2 node) 2M 1943 716 63.15% This patch (of 8): The readability of `hugetlb_hstate_alloc_pages` is poor. By cleaning the code, its readability can be improved, facilitating future modifications. This patch extracts two functions to reduce the complexity of `hugetlb_hstate_alloc_pages` and has no functional changes. - hugetlb_hstate_alloc_pages_node_specific() to handle iterates through each online node and performs allocation if necessary. - hugetlb_hstate_alloc_pages_report() report error during allocation. And the value of h->max_huge_pages is updated accordingly. Link: https://lkml.kernel.org/r/20240222140422.393911-1-gang.li@linux.dev Link: https://lkml.kernel.org/r/20240222140422.393911-2-gang.li@linux.dev Signed-off-by: Gang Li <ligang.bdlg@bytedance.com> Tested-by: David Rientjes <rientjes@google.com> Reviewed-by: Muchun Song <muchun.song@linux.dev> Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com> Cc: Daniel Jordan <daniel.m.jordan@oracle.com> Cc: David Hildenbrand <david@redhat.com> Cc: Jane Chu <jane.chu@oracle.com> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Steffen Klassert <steffen.klassert@secunet.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff fc37bbb3 Thu Feb 22 07:04:14 MST 2024 Gang Li <gang.li@linux.dev> hugetlb: code clean for hugetlb_hstate_alloc_pages Patch series "hugetlb: parallelize hugetlb page init on boot", v6. Introduction ------------ Hugetlb initialization during boot takes up a considerable amount of time. For instance, on a 2TB system, initializing 1,800 1GB huge pages takes 1-2 seconds out of 10 seconds. Initializing 11,776 1GB pages on a 12TB Intel host takes more than 1 minute[1]. This is a noteworthy figure. Inspired by [2] and [3], hugetlb initialization can also be accelerated through parallelization. Kernel already has infrastructure like padata_do_multithreaded, this patch uses it to achieve effective results by minimal modifications. [1] https://lore.kernel.org/all/783f8bac-55b8-5b95-eb6a-11a583675000@google.com/ [2] https://lore.kernel.org/all/20200527173608.2885243-1-daniel.m.jordan@oracle.com/ [3] https://lore.kernel.org/all/20230906112605.2286994-1-usama.arif@bytedance.com/ [4] https://lore.kernel.org/all/76becfc1-e609-e3e8-2966-4053143170b6@google.com/ max_threads ----------- This patch use `padata_do_multithreaded` like this: ``` job.max_threads = num_node_state(N_MEMORY) * multiplier; padata_do_multithreaded(&job); ``` To fully utilize the CPU, the number of parallel threads needs to be carefully considered. `max_threads = num_node_state(N_MEMORY)` does not fully utilize the CPU, so we need to multiply it by a multiplier. Tests below indicate that a multiplier of 2 significantly improves performance, and although larger values also provide improvements, the gains are marginal. multiplier 1 2 3 4 5 ------------ ------- ------- ------- ------- ------- 256G 2node 358ms 215ms 157ms 134ms 126ms 2T 4node 979ms 679ms 543ms 489ms 481ms 50G 2node 71ms 44ms 37ms 30ms 31ms Therefore, choosing 2 as the multiplier strikes a good balance between enhancing parallel processing capabilities and maintaining efficient resource management. Test result ----------- test case no patch(ms) patched(ms) saved ------------------- -------------- ------------- -------- 256c2T(4 node) 1G 4745 2024 57.34% 128c1T(2 node) 1G 3358 1712 49.02% 12T 1G 77000 18300 76.23% 256c2T(4 node) 2M 3336 1051 68.52% 128c1T(2 node) 2M 1943 716 63.15% This patch (of 8): The readability of `hugetlb_hstate_alloc_pages` is poor. By cleaning the code, its readability can be improved, facilitating future modifications. This patch extracts two functions to reduce the complexity of `hugetlb_hstate_alloc_pages` and has no functional changes. - hugetlb_hstate_alloc_pages_node_specific() to handle iterates through each online node and performs allocation if necessary. - hugetlb_hstate_alloc_pages_report() report error during allocation. And the value of h->max_huge_pages is updated accordingly. Link: https://lkml.kernel.org/r/20240222140422.393911-1-gang.li@linux.dev Link: https://lkml.kernel.org/r/20240222140422.393911-2-gang.li@linux.dev Signed-off-by: Gang Li <ligang.bdlg@bytedance.com> Tested-by: David Rientjes <rientjes@google.com> Reviewed-by: Muchun Song <muchun.song@linux.dev> Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com> Cc: Daniel Jordan <daniel.m.jordan@oracle.com> Cc: David Hildenbrand <david@redhat.com> Cc: Jane Chu <jane.chu@oracle.com> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Steffen Klassert <steffen.klassert@secunet.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff fc37bbb3 Thu Feb 22 07:04:14 MST 2024 Gang Li <gang.li@linux.dev> hugetlb: code clean for hugetlb_hstate_alloc_pages Patch series "hugetlb: parallelize hugetlb page init on boot", v6. Introduction ------------ Hugetlb initialization during boot takes up a considerable amount of time. For instance, on a 2TB system, initializing 1,800 1GB huge pages takes 1-2 seconds out of 10 seconds. Initializing 11,776 1GB pages on a 12TB Intel host takes more than 1 minute[1]. This is a noteworthy figure. Inspired by [2] and [3], hugetlb initialization can also be accelerated through parallelization. Kernel already has infrastructure like padata_do_multithreaded, this patch uses it to achieve effective results by minimal modifications. [1] https://lore.kernel.org/all/783f8bac-55b8-5b95-eb6a-11a583675000@google.com/ [2] https://lore.kernel.org/all/20200527173608.2885243-1-daniel.m.jordan@oracle.com/ [3] https://lore.kernel.org/all/20230906112605.2286994-1-usama.arif@bytedance.com/ [4] https://lore.kernel.org/all/76becfc1-e609-e3e8-2966-4053143170b6@google.com/ max_threads ----------- This patch use `padata_do_multithreaded` like this: ``` job.max_threads = num_node_state(N_MEMORY) * multiplier; padata_do_multithreaded(&job); ``` To fully utilize the CPU, the number of parallel threads needs to be carefully considered. `max_threads = num_node_state(N_MEMORY)` does not fully utilize the CPU, so we need to multiply it by a multiplier. Tests below indicate that a multiplier of 2 significantly improves performance, and although larger values also provide improvements, the gains are marginal. multiplier 1 2 3 4 5 ------------ ------- ------- ------- ------- ------- 256G 2node 358ms 215ms 157ms 134ms 126ms 2T 4node 979ms 679ms 543ms 489ms 481ms 50G 2node 71ms 44ms 37ms 30ms 31ms Therefore, choosing 2 as the multiplier strikes a good balance between enhancing parallel processing capabilities and maintaining efficient resource management. Test result ----------- test case no patch(ms) patched(ms) saved ------------------- -------------- ------------- -------- 256c2T(4 node) 1G 4745 2024 57.34% 128c1T(2 node) 1G 3358 1712 49.02% 12T 1G 77000 18300 76.23% 256c2T(4 node) 2M 3336 1051 68.52% 128c1T(2 node) 2M 1943 716 63.15% This patch (of 8): The readability of `hugetlb_hstate_alloc_pages` is poor. By cleaning the code, its readability can be improved, facilitating future modifications. This patch extracts two functions to reduce the complexity of `hugetlb_hstate_alloc_pages` and has no functional changes. - hugetlb_hstate_alloc_pages_node_specific() to handle iterates through each online node and performs allocation if necessary. - hugetlb_hstate_alloc_pages_report() report error during allocation. And the value of h->max_huge_pages is updated accordingly. Link: https://lkml.kernel.org/r/20240222140422.393911-1-gang.li@linux.dev Link: https://lkml.kernel.org/r/20240222140422.393911-2-gang.li@linux.dev Signed-off-by: Gang Li <ligang.bdlg@bytedance.com> Tested-by: David Rientjes <rientjes@google.com> Reviewed-by: Muchun Song <muchun.song@linux.dev> Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com> Cc: Daniel Jordan <daniel.m.jordan@oracle.com> Cc: David Hildenbrand <david@redhat.com> Cc: Jane Chu <jane.chu@oracle.com> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Steffen Klassert <steffen.klassert@secunet.com> Cc: Alexey Dobriyan <adobriyan@gmail.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 7dac0ec8 Wed Feb 21 16:47:30 MST 2024 Vishal Moola (Oracle) <vishal.moola@gmail.com> hugetlb: pass struct vm_fault through to hugetlb_handle_userfault() Now that hugetlb_fault() has a struct vm_fault, have hugetlb_handle_userfault() use it instead of creating one of its own. This lets us reduce the number of arguments passed to hugetlb_handle_userfault() from 7 to 3, cleaning up the code and stack. Link: https://lkml.kernel.org/r/20240221234732.187629-4-vishal.moola@gmail.com Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Muchun Song <muchun.song@linux.dev> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 0ca22723 Wed Feb 21 16:47:29 MST 2024 Vishal Moola (Oracle) <vishal.moola@gmail.com> hugetlb: move vm_fault declaration to the top of hugetlb_fault() hugetlb_fault() currently defines a vm_fault to pass to the generic handle_userfault() function. We can move this definition to the top of hugetlb_fault() so that it can be used throughout the rest of the hugetlb fault path. This will help cleanup a number of excess variables and function arguments throughout the stack. Also, since vm_fault already has space to store the page offset, use that instead and get rid of idx. Link: https://lkml.kernel.org/r/20240221234732.187629-3-vishal.moola@gmail.com Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Muchun Song <muchun.song@linux.dev> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 831bc31a Mon Feb 05 20:08:11 MST 2024 Baolin Wang <baolin.wang@linux.alibaba.com> mm: hugetlb: improve the handling of hugetlb allocation failure for freed or in-use hugetlb alloc_and_dissolve_hugetlb_folio() preallocates a new hugetlb page before it takes hugetlb_lock. In 3 out of 4 cases the page is not really used and therefore the newly allocated page is just freed right away. This is wasteful and it might cause pre-mature failures in those cases. Address that by moving the allocation down to the only case (hugetlb page is really in the free pages pool). We need to drop hugetlb_lock to do so and therefore need to recheck the page state after regaining it. The patch is more of a cleanup than an actual fix to an existing problem. There are no known reports about pre-mature failures. Link: https://lkml.kernel.org/r/62890fd60b1ecd5bf1cdc476c973f60fe37aa0cb.1707181934.git.baolin.wang@linux.alibaba.com Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com> Acked-by: Michal Hocko <mhocko@suse.com> Reviewed-by: Muchun Song <muchun.song@linux.dev> Cc: David Hildenbrand <david@redhat.com> Cc: Oscar Salvador <osalvador@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> |
H A D | nommu.c | diff b6b7a8fa Mon Jan 02 09:08:55 MST 2023 David Hildenbrand <david@redhat.com> mm/nommu: don't use VM_MAYSHARE for MAP_PRIVATE mappings Let's stop using VM_MAYSHARE for MAP_PRIVATE mappings and use VM_MAYOVERLAY instead. Rewrite determine_vm_flags() to make the whole logic easier to digest, and to cleanly separate MAP_PRIVATE vs. MAP_SHARED. No functional change intended. Link: https://lkml.kernel.org/r/20230102160856.500584-3-david@redhat.com Signed-off-by: David Hildenbrand <david@redhat.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Jens Axboe <axboe@kernel.dk> Cc: Nicolas Pitre <nico@fluxnic.net> Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff fc4f4be9 Mon Jan 02 09:08:54 MST 2023 David Hildenbrand <david@redhat.com> mm/nommu: factor out check for NOMMU shared mappings into is_nommu_shared_mapping() Patch series "mm/nommu: don't use VM_MAYSHARE for MAP_PRIVATE mappings". Trying to reduce the confusion around VM_SHARED and VM_MAYSHARE first requires !CONFIG_MMU to stop using VM_MAYSHARE for MAP_PRIVATE mappings. CONFIG_MMU only sets VM_MAYSHARE for MAP_SHARED mappings. This paves the way for further VM_MAYSHARE and VM_SHARED cleanups: for example, renaming VM_MAYSHARED to VM_MAP_SHARED to make it cleaner what is actually means. Let's first get the weird case out of the way and not use VM_MAYSHARE in MAP_PRIVATE mappings, using a new VM_MAYOVERLAY flag instead. This patch (of 3): We want to stop using VM_MAYSHARE in private mappings to pave the way for clarifying the semantics of VM_MAYSHARE vs. VM_SHARED and reduce the confusion. While CONFIG_MMU uses VM_MAYSHARE to represent MAP_SHARED, !CONFIG_MMU also sets VM_MAYSHARE for selected R/O private file mappings that are an effective overlay of a file mapping. Let's factor out all relevant VM_MAYSHARE checks in !CONFIG_MMU code into is_nommu_shared_mapping() first. Note that whenever VM_SHARED is set, VM_MAYSHARE must be set as well (unless there is a serious BUG). So there is not need to test for VM_SHARED manually. No functional change intended. Link: https://lkml.kernel.org/r/20230102160856.500584-1-david@redhat.com Link: https://lkml.kernel.org/r/20230102160856.500584-2-david@redhat.com Signed-off-by: David Hildenbrand <david@redhat.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: David Hildenbrand <david@redhat.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Jens Axboe <axboe@kernel.dk> Cc: Nicolas Pitre <nico@fluxnic.net> Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 3b8db39f Mon Jun 28 20:38:35 MDT 2021 David Hildenbrand <david@redhat.com> mm: ignore MAP_EXECUTABLE in ksys_mmap_pgoff() Let's also remove masking off MAP_EXECUTABLE from ksys_mmap_pgoff(): the last in-tree occurrence of MAP_EXECUTABLE is now in LEGACY_MAP_MASK, which accepts the flag e.g., for MAP_SHARED_VALIDATE; however, the flag is ignored throughout the kernel now. Add a comment to LEGACY_MAP_MASK stating that MAP_EXECUTABLE is ignored. Link: https://lkml.kernel.org/r/20210421093453.6904-4-david@redhat.com Signed-off-by: David Hildenbrand <david@redhat.com> Acked-by: "Eric W. Biederman" <ebiederm@xmission.com> Reviewed-by: Kees Cook <keescook@chromium.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Arnaldo Carvalho de Melo <acme@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Don Zickus <dzickus@redhat.com> Cc: Feng Tang <feng.tang@intel.com> Cc: Greg Ungerer <gerg@linux-m68k.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Kevin Brodsky <Kevin.Brodsky@arm.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Michal Hocko <mhocko@suse.com> Cc: Mike Rapoport <rppt@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 3f98a28c Thu Jan 28 03:06:26 MST 2021 Geert Uytterhoeven <geert@linux-m68k.org> mm/nommu: Fix return type of filemap_map_pages() If CONFIG_MMU is not set (e.g. m68k/m5272c3_defconfig): mm/nommu.c:1671:6: error: conflicting types for ‘filemap_map_pages’ 1671 | void filemap_map_pages(struct vm_fault *vmf, | ^~~~~~~~~~~~~~~~~ In file included from mm/nommu.c:20: ./include/linux/mm.h:2578:19: note: previous declaration of ‘filemap_map_pages’ was here 2578 | extern vm_fault_t filemap_map_pages(struct vm_fault *vmf, | ^~~~~~~~~~~~~~~~~ The signature of filemap_map_pages() was changed, but the nommu implementation wasn't updated. Reported-by: noreply@ellerman.id.au Fixes: f9ce0be71d1f ("mm: Cleanup faultaround and finish_fault() codepaths") Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org> Link: https://lore.kernel.org/r/20210128100626.2257638-1-geert@linux-m68k.org Signed-off-by: Will Deacon <will@kernel.org> diff 800c02f5 Tue Jun 23 07:31:36 MDT 2020 Mauro Carvalho Chehab <mchehab+huawei@kernel.org> docs: move nommu-mmap.txt to admin-guide and rename to ReST The nommu-mmap.txt file provides description of user visible behaviuour. So, move it to the admin-guide. As it is already at the ReST, also rename it. Suggested-by: Mike Rapoport <rppt@linux.ibm.com> Suggested-by: Jonathan Corbet <corbet@lwn.net> Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org> Link: https://lore.kernel.org/r/3a63d1833b513700755c85bf3bda0a6c4ab56986.1592918949.git.mchehab+huawei@kernel.org Signed-off-by: Jonathan Corbet <corbet@lwn.net> diff 763802b5 Sat Mar 21 19:22:41 MDT 2020 Joerg Roedel <jroedel@suse.de> x86/mm: split vmalloc_sync_all() Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in the vunmap() code-path. While this change was necessary to maintain correctness on x86-32-pae kernels, it also adds additional cycles for architectures that don't need it. Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported severe performance regressions in micro-benchmarks because it now also calls the x86-64 implementation of vmalloc_sync_all() on vunmap(). But the vmalloc_sync_all() implementation on x86-64 is only needed for newly created mappings. To avoid the unnecessary work on x86-64 and to gain the performance back, split up vmalloc_sync_all() into two functions: * vmalloc_sync_mappings(), and * vmalloc_sync_unmappings() Most call-sites to vmalloc_sync_all() only care about new mappings being synchronized. The only exception is the new call-site added in the above mentioned commit. Shile Zhang directed us to a report of an 80% regression in reaim throughput. Fixes: 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") Reported-by: kernel test robot <oliver.sang@intel.com> Reported-by: Shile Zhang <shile.zhang@linux.alibaba.com> Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Borislav Petkov <bp@suse.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> [GHES] Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: <stable@vger.kernel.org> Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2KEPC6KGKS6J25AIDB/ Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 763802b5 Sat Mar 21 19:22:41 MDT 2020 Joerg Roedel <jroedel@suse.de> x86/mm: split vmalloc_sync_all() Commit 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") introduced a call to vmalloc_sync_all() in the vunmap() code-path. While this change was necessary to maintain correctness on x86-32-pae kernels, it also adds additional cycles for architectures that don't need it. Specifically on x86-64 with CONFIG_VMAP_STACK=y some people reported severe performance regressions in micro-benchmarks because it now also calls the x86-64 implementation of vmalloc_sync_all() on vunmap(). But the vmalloc_sync_all() implementation on x86-64 is only needed for newly created mappings. To avoid the unnecessary work on x86-64 and to gain the performance back, split up vmalloc_sync_all() into two functions: * vmalloc_sync_mappings(), and * vmalloc_sync_unmappings() Most call-sites to vmalloc_sync_all() only care about new mappings being synchronized. The only exception is the new call-site added in the above mentioned commit. Shile Zhang directed us to a report of an 80% regression in reaim throughput. Fixes: 3f8fd02b1bf1 ("mm/vmalloc: Sync unmappings in __purge_vmap_area_lazy()") Reported-by: kernel test robot <oliver.sang@intel.com> Reported-by: Shile Zhang <shile.zhang@linux.alibaba.com> Signed-off-by: Joerg Roedel <jroedel@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Borislav Petkov <bp@suse.de> Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> [GHES] Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: <stable@vger.kernel.org> Link: http://lkml.kernel.org/r/20191009124418.8286-1-joro@8bytes.org Link: https://lists.01.org/hyperkitty/list/lkp@lists.01.org/thread/4D3JPPHBNOSPFK2KEPC6KGKS6J25AIDB/ Link: http://lkml.kernel.org/r/20191113095530.228959-1-shile.zhang@linux.alibaba.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 1b9fc5b2 Sat Nov 30 18:50:49 MST 2019 Wei Yang <richardw.yang@linux.intel.com> mm/mmap.c: extract __vma_unlink_list() as counterpart for __vma_link_list() Just make the code a little easier to read. Link: http://lkml.kernel.org/r/20191006012636.31521-3-richardw.yang@linux.intel.com Signed-off-by: Wei Yang <richardw.yang@linux.intel.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Oscar Salvador <osalvador@suse.de> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff a50b854e Mon Sep 23 16:34:25 MDT 2019 Matthew Wilcox (Oracle) <willy@infradead.org> mm: introduce page_size() Patch series "Make working with compound pages easier", v2. These three patches add three helpers and convert the appropriate places to use them. This patch (of 3): It's unnecessarily hard to find out the size of a potentially huge page. Replace 'PAGE_SIZE << compound_order(page)' with page_size(page). Link: http://lkml.kernel.org/r/20190721104612.19120-2-willy@infradead.org Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Acked-by: Michal Hocko <mhocko@suse.com> Reviewed-by: Andrew Morton <akpm@linux-foundation.org> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff df06b37f Fri Oct 26 16:10:28 MDT 2018 Keith Busch <kbusch@kernel.org> mm/gup: cache dev_pagemap while pinning pages Getting pages from ZONE_DEVICE memory needs to check the backing device's live-ness, which is tracked in the device's dev_pagemap metadata. This metadata is stored in a radix tree and looking it up adds measurable software overhead. This patch avoids repeating this relatively costly operation when dev_pagemap is used by caching the last dev_pagemap while getting user pages. The gup_benchmark kernel self test reports this reduces time to get user pages to as low as 1/3 of the previous time. Link: http://lkml.kernel.org/r/20181012173040.15669-1-keith.busch@intel.com Signed-off-by: Keith Busch <keith.busch@intel.com> Reviewed-by: Dan Williams <dan.j.williams@intel.com> Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> |
H A D | sparse.c | diff c5f1e2d1 Mon Jan 08 06:27:43 MST 2024 Sumanth Korikkar <sumanthk@linux.ibm.com> mm/memory_hotplug: introduce MEM_PREPARE_ONLINE/MEM_FINISH_OFFLINE notifiers Patch series "implement "memmap on memory" feature on s390". This series provides "memmap on memory" support on s390 platform. "memmap on memory" allows struct pages array to be allocated from the hotplugged memory range instead of allocating it from main system memory. s390 currently preallocates struct pages array for all potentially possible memory, which ensures memory onlining always succeeds, but with the cost of significant memory consumption from the available system memory during boottime. In certain extreme configuration, this could lead to ipl failure. "memmap on memory" ensures struct pages array are populated from self contained hotplugged memory range instead of depleting the available system memory and this could eliminate ipl failure on s390 platform. On other platforms, system might go OOM when the physically hotplugged memory depletes the available memory before it is onlined. Hence, "memmap on memory" feature was introduced as described in commit a08a2ae34613 ("mm,memory_hotplug: allocate memmap from the added memory range"). Unlike other architectures, s390 memory blocks are not physically accessible until it is online. To make it physically accessible two new memory notifiers MEM_PREPARE_ONLINE / MEM_FINISH_OFFLINE are added and this notifier lets the hypervisor inform that the memory should be made physically accessible. This allows for "memmap on memory" initialization during memory hotplug onlining phase, which is performed before calling MEM_GOING_ONLINE notifier. Patch 1 introduces MEM_PREPARE_ONLINE/MEM_FINISH_OFFLINE memory notifiers to prepare the transition of memory to and from a physically accessible state. New mhp_flag MHP_OFFLINE_INACCESSIBLE is introduced to ensure altmap cannot be written when adding memory - before it is set online. This enhancement is crucial for implementing the "memmap on memory" feature for s390 in a subsequent patch. Patches 2 allocates vmemmap pages from self-contained memory range for s390. It allocates memory map (struct pages array) from the hotplugged memory range, rather than using system memory by passing altmap to vmemmap functions. Patch 3 removes unhandled memory notifier types on s390. Patch 4 implements MEM_PREPARE_ONLINE/MEM_FINISH_OFFLINE memory notifiers on s390. MEM_PREPARE_ONLINE memory notifier makes memory block physical accessible via sclp assign command. The notifier ensures self-contained memory maps are accessible and hence enabling the "memmap on memory" on s390. MEM_FINISH_OFFLINE memory notifier shifts the memory block to an inaccessible state via sclp unassign command. Patch 5 finally enables MHP_MEMMAP_ON_MEMORY on s390. This patch (of 5): Introduce MEM_PREPARE_ONLINE/MEM_FINISH_OFFLINE memory notifiers to prepare the transition of memory to and from a physically accessible state. This enhancement is crucial for implementing the "memmap on memory" feature for s390 in a subsequent patch. Platforms such as x86 can support physical memory hotplug via ACPI. When there is physical memory hotplug, ACPI event leads to the memory addition with the following callchain: acpi_memory_device_add() -> acpi_memory_enable_device() -> __add_memory() After this, the hotplugged memory is physically accessible, and altmap support prepared, before the "memmap on memory" initialization in memory_block_online() is called. On s390, memory hotplug works in a different way. The available hotplug memory has to be defined upfront in the hypervisor, but it is made physically accessible only when the user sets it online via sysfs, currently in the MEM_GOING_ONLINE notifier. This is too late and "memmap on memory" initialization is performed before calling MEM_GOING_ONLINE notifier. During the memory hotplug addition phase, altmap support is prepared and during the memory onlining phase s390 requires memory to be physically accessible and then subsequently initiate the "memmap on memory" initialization process. The memory provider will handle new MEM_PREPARE_ONLINE / MEM_FINISH_OFFLINE notifications and make the memory accessible. The mhp_flag MHP_OFFLINE_INACCESSIBLE is introduced and is relevant when used along with MHP_MEMMAP_ON_MEMORY, because the altmap cannot be written (e.g., poisoned) when adding memory -- before it is set online. This allows for adding memory with an altmap that is not currently made available by a hypervisor. When onlining that memory, the hypervisor can be instructed to make that memory accessible via the new notifiers and the onlining phase will not require any memory allocations, which is helpful in low-memory situations. All architectures ignore unknown memory notifiers. Therefore, the introduction of these new notifiers does not result in any functional modifications across architectures. Link: https://lkml.kernel.org/r/20240108132747.3238763-1-sumanthk@linux.ibm.com Link: https://lkml.kernel.org/r/20240108132747.3238763-2-sumanthk@linux.ibm.com Signed-off-by: Sumanth Korikkar <sumanthk@linux.ibm.com> Suggested-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com> Suggested-by: David Hildenbrand <david@redhat.com> Acked-by: David Hildenbrand <david@redhat.com> Cc: Alexander Gordeev <agordeev@linux.ibm.com> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> Cc: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Heiko Carstens <hca@linux.ibm.com> Cc: Michal Hocko <mhocko@suse.com> Cc: Oscar Salvador <osalvador@suse.de> Cc: Vasily Gorbik <gor@linux.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 60f272f6 Thu May 12 21:23:09 MDT 2022 zhenwei pi <pizhenwei@bytedance.com> mm/memory-failure.c: move clear_hwpoisoned_pages Patch series "memory-failure: fix hwpoison_filter", v2. As well known, the memory failure mechanism handles memory corrupted event, and try to send SIGBUS to the user process which uses this corrupted page. For the virtualization case, QEMU catches SIGBUS and tries to inject MCE into the guest, and the guest handles memory failure again. Thus the guest gets the minimal effect from hardware memory corruption. The further step I'm working on: 1, try to modify code to decrease poisoned pages in a single place (mm/memofy-failure.c: simplify num_poisoned_pages_dec in this series). 2, try to use page_handle_poison() to handle SetPageHWPoison() and num_poisoned_pages_inc() together. It would be best to call num_poisoned_pages_inc() in a single place too. 3, introduce memory failure notifier list in memory-failure.c: notify the corrupted PFN to someone who registers this list. If I can complete [1] and [2] part, [3] will be quite easy(just call notifier list after increasing poisoned page). 4, introduce memory recover VQ for memory balloon device, and registers memory failure notifier list. During the guest kernel handles memory failure, balloon device gets notified by memory failure notifier list, and tells the host to recover the corrupted PFN(GPA) by the new VQ. 5, host side remaps the corrupted page(HVA), and tells the guest side to unpoison the PFN(GPA). Then the guest fixes the corrupted page(GPA) dynamically. This patch (of 5): clear_hwpoisoned_pages() clears HWPoison flag and decreases the number of poisoned pages, this actually works as part of memory failure. Move this function from sparse.c to memory-failure.c, finally there is no CONFIG_MEMORY_FAILURE in sparse.c. Link: https://lkml.kernel.org/r/20220509105641.491313-1-pizhenwei@bytedance.com Link: https://lkml.kernel.org/r/20220509105641.491313-2-pizhenwei@bytedance.com Signed-off-by: zhenwei pi <pizhenwei@bytedance.com> Acked-by: Naoya Horiguchi <naoya.horiguchi@nec.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 60f272f6 Thu May 12 21:23:09 MDT 2022 zhenwei pi <pizhenwei@bytedance.com> mm/memory-failure.c: move clear_hwpoisoned_pages Patch series "memory-failure: fix hwpoison_filter", v2. As well known, the memory failure mechanism handles memory corrupted event, and try to send SIGBUS to the user process which uses this corrupted page. For the virtualization case, QEMU catches SIGBUS and tries to inject MCE into the guest, and the guest handles memory failure again. Thus the guest gets the minimal effect from hardware memory corruption. The further step I'm working on: 1, try to modify code to decrease poisoned pages in a single place (mm/memofy-failure.c: simplify num_poisoned_pages_dec in this series). 2, try to use page_handle_poison() to handle SetPageHWPoison() and num_poisoned_pages_inc() together. It would be best to call num_poisoned_pages_inc() in a single place too. 3, introduce memory failure notifier list in memory-failure.c: notify the corrupted PFN to someone who registers this list. If I can complete [1] and [2] part, [3] will be quite easy(just call notifier list after increasing poisoned page). 4, introduce memory recover VQ for memory balloon device, and registers memory failure notifier list. During the guest kernel handles memory failure, balloon device gets notified by memory failure notifier list, and tells the host to recover the corrupted PFN(GPA) by the new VQ. 5, host side remaps the corrupted page(HVA), and tells the guest side to unpoison the PFN(GPA). Then the guest fixes the corrupted page(GPA) dynamically. This patch (of 5): clear_hwpoisoned_pages() clears HWPoison flag and decreases the number of poisoned pages, this actually works as part of memory failure. Move this function from sparse.c to memory-failure.c, finally there is no CONFIG_MEMORY_FAILURE in sparse.c. Link: https://lkml.kernel.org/r/20220509105641.491313-1-pizhenwei@bytedance.com Link: https://lkml.kernel.org/r/20220509105641.491313-2-pizhenwei@bytedance.com Signed-off-by: zhenwei pi <pizhenwei@bytedance.com> Acked-by: Naoya Horiguchi <naoya.horiguchi@nec.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff e3246d8f Fri Apr 29 00:16:15 MDT 2022 Joao Martins <joao.m.martins@oracle.com> mm/sparse-vmemmap: add a pgmap argument to section activation Patch series "sparse-vmemmap: memory savings for compound devmaps (device-dax)", v9. This series minimizes 'struct page' overhead by pursuing a similar approach as Muchun Song series "Free some vmemmap pages of hugetlb page" (now merged since v5.14), but applied to devmap with @vmemmap_shift (device-dax). The vmemmap dedpulication original idea (already used in HugeTLB) is to reuse/deduplicate tail page vmemmap areas, particular the area which only describes tail pages. So a vmemmap page describes 64 struct pages, and the first page for a given ZONE_DEVICE vmemmap would contain the head page and 63 tail pages. The second vmemmap page would contain only tail pages, and that's what gets reused across the rest of the subsection/section. The bigger the page size, the bigger the savings (2M hpage -> save 6 vmemmap pages; 1G hpage -> save 4094 vmemmap pages). This is done for PMEM /specifically only/ on device-dax configured namespaces, not fsdax. In other words, a devmap with a @vmemmap_shift. In terms of savings, per 1Tb of memory, the struct page cost would go down with compound devmap: * with 2M pages we lose 4G instead of 16G (0.39% instead of 1.5% of total memory) * with 1G pages we lose 40MB instead of 16G (0.0014% instead of 1.5% of total memory) The series is mostly summed up by patch 4, and to summarize what the series does: Patches 1 - 3: Minor cleanups in preparation for patch 4. Move the very nice docs of hugetlb_vmemmap.c into a Documentation/vm/ entry. Patch 4: Patch 4 is the one that takes care of the struct page savings (also referred to here as tail-page/vmemmap deduplication). Much like Muchun series, we reuse the second PTE tail page vmemmap areas across a given @vmemmap_shift On important difference though, is that contrary to the hugetlbfs series, there's no vmemmap for the area because we are late-populating it as opposed to remapping a system-ram range. IOW no freeing of pages of already initialized vmemmap like the case for hugetlbfs, which greatly simplifies the logic (besides not being arch-specific). altmap case unchanged and still goes via the vmemmap_populate(). Also adjust the newly added docs to the device-dax case. [Note that device-dax is still a little behind HugeTLB in terms of savings. I have an additional simple patch that reuses the head vmemmap page too, as a follow-up. That will double the savings and namespaces initialization.] Patch 5: Initialize fewer struct pages depending on the page size with DRAM backed struct pages -- because fewer pages are unique and most tail pages (with bigger vmemmap_shift). NVDIMM namespace bootstrap improves from ~268-358 ms to ~80-110/<1ms on 128G NVDIMMs with 2M and 1G respectivally. And struct page needed capacity will be 3.8x / 1071x smaller for 2M and 1G respectivelly. Tested on x86 with 1.5Tb of pmem (including pinning, and RDMA registration/deregistration scalability with 2M MRs) This patch (of 5): In support of using compound pages for devmap mappings, plumb the pgmap down to the vmemmap_populate implementation. Note that while altmap is retrievable from pgmap the memory hotplug code passes altmap without pgmap[*], so both need to be independently plumbed. So in addition to @altmap, pass @pgmap to sparse section populate functions namely: sparse_add_section section_activate populate_section_memmap __populate_section_memmap Passing @pgmap allows __populate_section_memmap() to both fetch the vmemmap_shift in which memmap metadata is created for and also to let sparse-vmemmap fetch pgmap ranges to co-relate to a given section and pick whether to just reuse tail pages from past onlined sections. While at it, fix the kdoc for @altmap for sparse_add_section(). [*] https://lore.kernel.org/linux-mm/20210319092635.6214-1-osalvador@suse.de/ Link: https://lkml.kernel.org/r/20220420155310.9712-1-joao.m.martins@oracle.com Link: https://lkml.kernel.org/r/20220420155310.9712-2-joao.m.martins@oracle.com Signed-off-by: Joao Martins <joao.m.martins@oracle.com> Reviewed-by: Dan Williams <dan.j.williams@intel.com> Reviewed-by: Muchun Song <songmuchun@bytedance.com> Cc: Vishal Verma <vishal.l.verma@intel.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Jason Gunthorpe <jgg@ziepe.ca> Cc: Jane Chu <jane.chu@oracle.com> Cc: Mike Kravetz <mike.kravetz@oracle.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 3ecc6834 Fri Nov 05 14:43:19 MDT 2021 Mike Rapoport <rppt@kernel.org> memblock: rename memblock_free to memblock_phys_free Since memblock_free() operates on a physical range, make its name reflect it and rename it to memblock_phys_free(), so it will be a logical counterpart to memblock_phys_alloc(). The callers are updated with the below semantic patch: @@ expression addr; expression size; @@ - memblock_free(addr, size); + memblock_phys_free(addr, size); Link: https://lkml.kernel.org/r/20210930185031.18648-6-rppt@kernel.org Signed-off-by: Mike Rapoport <rppt@linux.ibm.com> Cc: Christophe Leroy <christophe.leroy@csgroup.eu> Cc: Juergen Gross <jgross@suse.com> Cc: Shahab Vahedi <Shahab.Vahedi@synopsys.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff a1bc561b Thu Sep 02 15:56:58 MDT 2021 Ohhoon Kwon <ohoono.kwon@samsung.com> mm: sparse: pass section_nr to section_mark_present Patch series "mm: sparse: remove __section_nr() function", v4. This patch (of 3): With CONFIG_SPARSEMEM_EXTREME enabled, __section_nr() which converts mem_section to section_nr could be costly since it iterates all section roots to check if the given mem_section is in its range. Since both callers of section_mark_present already know section_nr, let's also pass section_nr as well as mem_section in order to reduce costly translation. Link: https://lkml.kernel.org/r/20210707150212.855-1-ohoono.kwon@samsung.com Link: https://lkml.kernel.org/r/20210707150212.855-2-ohoono.kwon@samsung.com Signed-off-by: Ohhoon Kwon <ohoono.kwon@samsung.com> Acked-by: Mike Rapoport <rppt@linux.ibm.com> Acked-by: Michal Hocko <mhocko@suse.com> Reviewed-by: David Hildenbrand <david@redhat.com> Cc: Baoquan He <bhe@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 426e5c42 Wed Jun 30 19:47:00 MDT 2021 Muchun Song <songmuchun@bytedance.com> mm: memory_hotplug: factor out bootmem core functions to bootmem_info.c Patch series "Free some vmemmap pages of HugeTLB page", v23. This patch series will free some vmemmap pages(struct page structures) associated with each HugeTLB page when preallocated to save memory. In order to reduce the difficulty of the first version of code review. In this version, we disable PMD/huge page mapping of vmemmap if this feature was enabled. This acutely eliminates a bunch of the complex code doing page table manipulation. When this patch series is solid, we cam add the code of vmemmap page table manipulation in the future. The struct page structures (page structs) are used to describe a physical page frame. By default, there is an one-to-one mapping from a page frame to it's corresponding page struct. The HugeTLB pages consist of multiple base page size pages and is supported by many architectures. See hugetlbpage.rst in the Documentation directory for more details. On the x86 architecture, HugeTLB pages of size 2MB and 1GB are currently supported. Since the base page size on x86 is 4KB, a 2MB HugeTLB page consists of 512 base pages and a 1GB HugeTLB page consists of 4096 base pages. For each base page, there is a corresponding page struct. Within the HugeTLB subsystem, only the first 4 page structs are used to contain unique information about a HugeTLB page. HUGETLB_CGROUP_MIN_ORDER provides this upper limit. The only 'useful' information in the remaining page structs is the compound_head field, and this field is the same for all tail pages. By removing redundant page structs for HugeTLB pages, memory can returned to the buddy allocator for other uses. When the system boot up, every 2M HugeTLB has 512 struct page structs which size is 8 pages(sizeof(struct page) * 512 / PAGE_SIZE). HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | -------------> | 2 | | | +-----------+ +-----------+ | | | 3 | -------------> | 3 | | | +-----------+ +-----------+ | | | 4 | -------------> | 4 | | 2MB | +-----------+ +-----------+ | | | 5 | -------------> | 5 | | | +-----------+ +-----------+ | | | 6 | -------------> | 6 | | | +-----------+ +-----------+ | | | 7 | -------------> | 7 | | | +-----------+ +-----------+ | | | | | | +-----------+ The value of page->compound_head is the same for all tail pages. The first page of page structs (page 0) associated with the HugeTLB page contains the 4 page structs necessary to describe the HugeTLB. The only use of the remaining pages of page structs (page 1 to page 7) is to point to page->compound_head. Therefore, we can remap pages 2 to 7 to page 1. Only 2 pages of page structs will be used for each HugeTLB page. This will allow us to free the remaining 6 pages to the buddy allocator. Here is how things look after remapping. HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | ----------------^ ^ ^ ^ ^ ^ | | +-----------+ | | | | | | | | 3 | ------------------+ | | | | | | +-----------+ | | | | | | | 4 | --------------------+ | | | | 2MB | +-----------+ | | | | | | 5 | ----------------------+ | | | | +-----------+ | | | | | 6 | ------------------------+ | | | +-----------+ | | | | 7 | --------------------------+ | | +-----------+ | | | | | | +-----------+ When a HugeTLB is freed to the buddy system, we should allocate 6 pages for vmemmap pages and restore the previous mapping relationship. Apart from 2MB HugeTLB page, we also have 1GB HugeTLB page. It is similar to the 2MB HugeTLB page. We also can use this approach to free the vmemmap pages. In this case, for the 1GB HugeTLB page, we can save 4094 pages. This is a very substantial gain. On our server, run some SPDK/QEMU applications which will use 1024GB HugeTLB page. With this feature enabled, we can save ~16GB (1G hugepage)/~12GB (2MB hugepage) memory. Because there are vmemmap page tables reconstruction on the freeing/allocating path, it increases some overhead. Here are some overhead analysis. 1) Allocating 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.166s user 0m0.000s sys 0m0.166s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 5476 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [16K, 32K) 4760 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ | [32K, 64K) 4 | | b) Without this patch series: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.067s user 0m0.000s sys 0m0.067s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 10147 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 93 | | Summarize: this feature is about ~2x slower than before. 2) Freeing 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.213s user 0m0.000s sys 0m0.213s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 6 | | [16K, 32K) 10227 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [32K, 64K) 7 | | b) Without this patch series: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.081s user 0m0.000s sys 0m0.081s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 6805 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 3427 |@@@@@@@@@@@@@@@@@@@@@@@@@@ | [16K, 32K) 8 | | Summary: The overhead of __free_hugepage is about ~2-3x slower than before. Although the overhead has increased, the overhead is not significant. Like Mike said, "However, remember that the majority of use cases create HugeTLB pages at or shortly after boot time and add them to the pool. So, additional overhead is at pool creation time. There is no change to 'normal run time' operations of getting a page from or returning a page to the pool (think page fault/unmap)". Despite the overhead and in addition to the memory gains from this series. The following data is obtained by Joao Martins. Very thanks to his effort. There's an additional benefit which is page (un)pinners will see an improvement and Joao presumes because there are fewer memmap pages and thus the tail/head pages are staying in cache more often. Out of the box Joao saw (when comparing linux-next against linux-next + this series) with gup_test and pinning a 16G HugeTLB file (with 1G pages): get_user_pages(): ~32k -> ~9k unpin_user_pages(): ~75k -> ~70k Usually any tight loop fetching compound_head(), or reading tail pages data (e.g. compound_head) benefit a lot. There's some unpinning inefficiencies Joao was fixing[2], but with that in added it shows even more: unpin_user_pages(): ~27k -> ~3.8k [1] https://lore.kernel.org/linux-mm/20210409205254.242291-1-mike.kravetz@oracle.com/ [2] https://lore.kernel.org/linux-mm/20210204202500.26474-1-joao.m.martins@oracle.com/ This patch (of 9): Move bootmem info registration common API to individual bootmem_info.c. And we will use {get,put}_page_bootmem() to initialize the page for the vmemmap pages or free the vmemmap pages to buddy in the later patch. So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code movement without any functional change. Link: https://lkml.kernel.org/r/20210510030027.56044-1-songmuchun@bytedance.com Link: https://lkml.kernel.org/r/20210510030027.56044-2-songmuchun@bytedance.com Signed-off-by: Muchun Song <songmuchun@bytedance.com> Acked-by: Mike Kravetz <mike.kravetz@oracle.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Reviewed-by: David Hildenbrand <david@redhat.com> Reviewed-by: Miaohe Lin <linmiaohe@huawei.com> Tested-by: Chen Huang <chenhuang5@huawei.com> Tested-by: Bodeddula Balasubramaniam <bodeddub@amazon.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: x86@kernel.org Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Pawan Gupta <pawan.kumar.gupta@linux.intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Oliver Neukum <oneukum@suse.com> Cc: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Joerg Roedel <jroedel@suse.de> Cc: Mina Almasry <almasrymina@google.com> Cc: David Rientjes <rientjes@google.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Barry Song <song.bao.hua@hisilicon.com> Cc: HORIGUCHI NAOYA <naoya.horiguchi@nec.com> Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Xiongchun Duan <duanxiongchun@bytedance.com> Cc: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 426e5c42 Wed Jun 30 19:47:00 MDT 2021 Muchun Song <songmuchun@bytedance.com> mm: memory_hotplug: factor out bootmem core functions to bootmem_info.c Patch series "Free some vmemmap pages of HugeTLB page", v23. This patch series will free some vmemmap pages(struct page structures) associated with each HugeTLB page when preallocated to save memory. In order to reduce the difficulty of the first version of code review. In this version, we disable PMD/huge page mapping of vmemmap if this feature was enabled. This acutely eliminates a bunch of the complex code doing page table manipulation. When this patch series is solid, we cam add the code of vmemmap page table manipulation in the future. The struct page structures (page structs) are used to describe a physical page frame. By default, there is an one-to-one mapping from a page frame to it's corresponding page struct. The HugeTLB pages consist of multiple base page size pages and is supported by many architectures. See hugetlbpage.rst in the Documentation directory for more details. On the x86 architecture, HugeTLB pages of size 2MB and 1GB are currently supported. Since the base page size on x86 is 4KB, a 2MB HugeTLB page consists of 512 base pages and a 1GB HugeTLB page consists of 4096 base pages. For each base page, there is a corresponding page struct. Within the HugeTLB subsystem, only the first 4 page structs are used to contain unique information about a HugeTLB page. HUGETLB_CGROUP_MIN_ORDER provides this upper limit. The only 'useful' information in the remaining page structs is the compound_head field, and this field is the same for all tail pages. By removing redundant page structs for HugeTLB pages, memory can returned to the buddy allocator for other uses. When the system boot up, every 2M HugeTLB has 512 struct page structs which size is 8 pages(sizeof(struct page) * 512 / PAGE_SIZE). HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | -------------> | 2 | | | +-----------+ +-----------+ | | | 3 | -------------> | 3 | | | +-----------+ +-----------+ | | | 4 | -------------> | 4 | | 2MB | +-----------+ +-----------+ | | | 5 | -------------> | 5 | | | +-----------+ +-----------+ | | | 6 | -------------> | 6 | | | +-----------+ +-----------+ | | | 7 | -------------> | 7 | | | +-----------+ +-----------+ | | | | | | +-----------+ The value of page->compound_head is the same for all tail pages. The first page of page structs (page 0) associated with the HugeTLB page contains the 4 page structs necessary to describe the HugeTLB. The only use of the remaining pages of page structs (page 1 to page 7) is to point to page->compound_head. Therefore, we can remap pages 2 to 7 to page 1. Only 2 pages of page structs will be used for each HugeTLB page. This will allow us to free the remaining 6 pages to the buddy allocator. Here is how things look after remapping. HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | ----------------^ ^ ^ ^ ^ ^ | | +-----------+ | | | | | | | | 3 | ------------------+ | | | | | | +-----------+ | | | | | | | 4 | --------------------+ | | | | 2MB | +-----------+ | | | | | | 5 | ----------------------+ | | | | +-----------+ | | | | | 6 | ------------------------+ | | | +-----------+ | | | | 7 | --------------------------+ | | +-----------+ | | | | | | +-----------+ When a HugeTLB is freed to the buddy system, we should allocate 6 pages for vmemmap pages and restore the previous mapping relationship. Apart from 2MB HugeTLB page, we also have 1GB HugeTLB page. It is similar to the 2MB HugeTLB page. We also can use this approach to free the vmemmap pages. In this case, for the 1GB HugeTLB page, we can save 4094 pages. This is a very substantial gain. On our server, run some SPDK/QEMU applications which will use 1024GB HugeTLB page. With this feature enabled, we can save ~16GB (1G hugepage)/~12GB (2MB hugepage) memory. Because there are vmemmap page tables reconstruction on the freeing/allocating path, it increases some overhead. Here are some overhead analysis. 1) Allocating 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.166s user 0m0.000s sys 0m0.166s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 5476 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [16K, 32K) 4760 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ | [32K, 64K) 4 | | b) Without this patch series: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.067s user 0m0.000s sys 0m0.067s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 10147 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 93 | | Summarize: this feature is about ~2x slower than before. 2) Freeing 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.213s user 0m0.000s sys 0m0.213s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 6 | | [16K, 32K) 10227 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [32K, 64K) 7 | | b) Without this patch series: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.081s user 0m0.000s sys 0m0.081s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 6805 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 3427 |@@@@@@@@@@@@@@@@@@@@@@@@@@ | [16K, 32K) 8 | | Summary: The overhead of __free_hugepage is about ~2-3x slower than before. Although the overhead has increased, the overhead is not significant. Like Mike said, "However, remember that the majority of use cases create HugeTLB pages at or shortly after boot time and add them to the pool. So, additional overhead is at pool creation time. There is no change to 'normal run time' operations of getting a page from or returning a page to the pool (think page fault/unmap)". Despite the overhead and in addition to the memory gains from this series. The following data is obtained by Joao Martins. Very thanks to his effort. There's an additional benefit which is page (un)pinners will see an improvement and Joao presumes because there are fewer memmap pages and thus the tail/head pages are staying in cache more often. Out of the box Joao saw (when comparing linux-next against linux-next + this series) with gup_test and pinning a 16G HugeTLB file (with 1G pages): get_user_pages(): ~32k -> ~9k unpin_user_pages(): ~75k -> ~70k Usually any tight loop fetching compound_head(), or reading tail pages data (e.g. compound_head) benefit a lot. There's some unpinning inefficiencies Joao was fixing[2], but with that in added it shows even more: unpin_user_pages(): ~27k -> ~3.8k [1] https://lore.kernel.org/linux-mm/20210409205254.242291-1-mike.kravetz@oracle.com/ [2] https://lore.kernel.org/linux-mm/20210204202500.26474-1-joao.m.martins@oracle.com/ This patch (of 9): Move bootmem info registration common API to individual bootmem_info.c. And we will use {get,put}_page_bootmem() to initialize the page for the vmemmap pages or free the vmemmap pages to buddy in the later patch. So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code movement without any functional change. Link: https://lkml.kernel.org/r/20210510030027.56044-1-songmuchun@bytedance.com Link: https://lkml.kernel.org/r/20210510030027.56044-2-songmuchun@bytedance.com Signed-off-by: Muchun Song <songmuchun@bytedance.com> Acked-by: Mike Kravetz <mike.kravetz@oracle.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Reviewed-by: David Hildenbrand <david@redhat.com> Reviewed-by: Miaohe Lin <linmiaohe@huawei.com> Tested-by: Chen Huang <chenhuang5@huawei.com> Tested-by: Bodeddula Balasubramaniam <bodeddub@amazon.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: x86@kernel.org Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Pawan Gupta <pawan.kumar.gupta@linux.intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Oliver Neukum <oneukum@suse.com> Cc: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Joerg Roedel <jroedel@suse.de> Cc: Mina Almasry <almasrymina@google.com> Cc: David Rientjes <rientjes@google.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Barry Song <song.bao.hua@hisilicon.com> Cc: HORIGUCHI NAOYA <naoya.horiguchi@nec.com> Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Xiongchun Duan <duanxiongchun@bytedance.com> Cc: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 426e5c42 Wed Jun 30 19:47:00 MDT 2021 Muchun Song <songmuchun@bytedance.com> mm: memory_hotplug: factor out bootmem core functions to bootmem_info.c Patch series "Free some vmemmap pages of HugeTLB page", v23. This patch series will free some vmemmap pages(struct page structures) associated with each HugeTLB page when preallocated to save memory. In order to reduce the difficulty of the first version of code review. In this version, we disable PMD/huge page mapping of vmemmap if this feature was enabled. This acutely eliminates a bunch of the complex code doing page table manipulation. When this patch series is solid, we cam add the code of vmemmap page table manipulation in the future. The struct page structures (page structs) are used to describe a physical page frame. By default, there is an one-to-one mapping from a page frame to it's corresponding page struct. The HugeTLB pages consist of multiple base page size pages and is supported by many architectures. See hugetlbpage.rst in the Documentation directory for more details. On the x86 architecture, HugeTLB pages of size 2MB and 1GB are currently supported. Since the base page size on x86 is 4KB, a 2MB HugeTLB page consists of 512 base pages and a 1GB HugeTLB page consists of 4096 base pages. For each base page, there is a corresponding page struct. Within the HugeTLB subsystem, only the first 4 page structs are used to contain unique information about a HugeTLB page. HUGETLB_CGROUP_MIN_ORDER provides this upper limit. The only 'useful' information in the remaining page structs is the compound_head field, and this field is the same for all tail pages. By removing redundant page structs for HugeTLB pages, memory can returned to the buddy allocator for other uses. When the system boot up, every 2M HugeTLB has 512 struct page structs which size is 8 pages(sizeof(struct page) * 512 / PAGE_SIZE). HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | -------------> | 2 | | | +-----------+ +-----------+ | | | 3 | -------------> | 3 | | | +-----------+ +-----------+ | | | 4 | -------------> | 4 | | 2MB | +-----------+ +-----------+ | | | 5 | -------------> | 5 | | | +-----------+ +-----------+ | | | 6 | -------------> | 6 | | | +-----------+ +-----------+ | | | 7 | -------------> | 7 | | | +-----------+ +-----------+ | | | | | | +-----------+ The value of page->compound_head is the same for all tail pages. The first page of page structs (page 0) associated with the HugeTLB page contains the 4 page structs necessary to describe the HugeTLB. The only use of the remaining pages of page structs (page 1 to page 7) is to point to page->compound_head. Therefore, we can remap pages 2 to 7 to page 1. Only 2 pages of page structs will be used for each HugeTLB page. This will allow us to free the remaining 6 pages to the buddy allocator. Here is how things look after remapping. HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | ----------------^ ^ ^ ^ ^ ^ | | +-----------+ | | | | | | | | 3 | ------------------+ | | | | | | +-----------+ | | | | | | | 4 | --------------------+ | | | | 2MB | +-----------+ | | | | | | 5 | ----------------------+ | | | | +-----------+ | | | | | 6 | ------------------------+ | | | +-----------+ | | | | 7 | --------------------------+ | | +-----------+ | | | | | | +-----------+ When a HugeTLB is freed to the buddy system, we should allocate 6 pages for vmemmap pages and restore the previous mapping relationship. Apart from 2MB HugeTLB page, we also have 1GB HugeTLB page. It is similar to the 2MB HugeTLB page. We also can use this approach to free the vmemmap pages. In this case, for the 1GB HugeTLB page, we can save 4094 pages. This is a very substantial gain. On our server, run some SPDK/QEMU applications which will use 1024GB HugeTLB page. With this feature enabled, we can save ~16GB (1G hugepage)/~12GB (2MB hugepage) memory. Because there are vmemmap page tables reconstruction on the freeing/allocating path, it increases some overhead. Here are some overhead analysis. 1) Allocating 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.166s user 0m0.000s sys 0m0.166s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 5476 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [16K, 32K) 4760 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ | [32K, 64K) 4 | | b) Without this patch series: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.067s user 0m0.000s sys 0m0.067s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 10147 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 93 | | Summarize: this feature is about ~2x slower than before. 2) Freeing 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.213s user 0m0.000s sys 0m0.213s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 6 | | [16K, 32K) 10227 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [32K, 64K) 7 | | b) Without this patch series: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.081s user 0m0.000s sys 0m0.081s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 6805 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 3427 |@@@@@@@@@@@@@@@@@@@@@@@@@@ | [16K, 32K) 8 | | Summary: The overhead of __free_hugepage is about ~2-3x slower than before. Although the overhead has increased, the overhead is not significant. Like Mike said, "However, remember that the majority of use cases create HugeTLB pages at or shortly after boot time and add them to the pool. So, additional overhead is at pool creation time. There is no change to 'normal run time' operations of getting a page from or returning a page to the pool (think page fault/unmap)". Despite the overhead and in addition to the memory gains from this series. The following data is obtained by Joao Martins. Very thanks to his effort. There's an additional benefit which is page (un)pinners will see an improvement and Joao presumes because there are fewer memmap pages and thus the tail/head pages are staying in cache more often. Out of the box Joao saw (when comparing linux-next against linux-next + this series) with gup_test and pinning a 16G HugeTLB file (with 1G pages): get_user_pages(): ~32k -> ~9k unpin_user_pages(): ~75k -> ~70k Usually any tight loop fetching compound_head(), or reading tail pages data (e.g. compound_head) benefit a lot. There's some unpinning inefficiencies Joao was fixing[2], but with that in added it shows even more: unpin_user_pages(): ~27k -> ~3.8k [1] https://lore.kernel.org/linux-mm/20210409205254.242291-1-mike.kravetz@oracle.com/ [2] https://lore.kernel.org/linux-mm/20210204202500.26474-1-joao.m.martins@oracle.com/ This patch (of 9): Move bootmem info registration common API to individual bootmem_info.c. And we will use {get,put}_page_bootmem() to initialize the page for the vmemmap pages or free the vmemmap pages to buddy in the later patch. So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code movement without any functional change. Link: https://lkml.kernel.org/r/20210510030027.56044-1-songmuchun@bytedance.com Link: https://lkml.kernel.org/r/20210510030027.56044-2-songmuchun@bytedance.com Signed-off-by: Muchun Song <songmuchun@bytedance.com> Acked-by: Mike Kravetz <mike.kravetz@oracle.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Reviewed-by: David Hildenbrand <david@redhat.com> Reviewed-by: Miaohe Lin <linmiaohe@huawei.com> Tested-by: Chen Huang <chenhuang5@huawei.com> Tested-by: Bodeddula Balasubramaniam <bodeddub@amazon.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: x86@kernel.org Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Pawan Gupta <pawan.kumar.gupta@linux.intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Oliver Neukum <oneukum@suse.com> Cc: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Joerg Roedel <jroedel@suse.de> Cc: Mina Almasry <almasrymina@google.com> Cc: David Rientjes <rientjes@google.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Barry Song <song.bao.hua@hisilicon.com> Cc: HORIGUCHI NAOYA <naoya.horiguchi@nec.com> Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Xiongchun Duan <duanxiongchun@bytedance.com> Cc: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 426e5c42 Wed Jun 30 19:47:00 MDT 2021 Muchun Song <songmuchun@bytedance.com> mm: memory_hotplug: factor out bootmem core functions to bootmem_info.c Patch series "Free some vmemmap pages of HugeTLB page", v23. This patch series will free some vmemmap pages(struct page structures) associated with each HugeTLB page when preallocated to save memory. In order to reduce the difficulty of the first version of code review. In this version, we disable PMD/huge page mapping of vmemmap if this feature was enabled. This acutely eliminates a bunch of the complex code doing page table manipulation. When this patch series is solid, we cam add the code of vmemmap page table manipulation in the future. The struct page structures (page structs) are used to describe a physical page frame. By default, there is an one-to-one mapping from a page frame to it's corresponding page struct. The HugeTLB pages consist of multiple base page size pages and is supported by many architectures. See hugetlbpage.rst in the Documentation directory for more details. On the x86 architecture, HugeTLB pages of size 2MB and 1GB are currently supported. Since the base page size on x86 is 4KB, a 2MB HugeTLB page consists of 512 base pages and a 1GB HugeTLB page consists of 4096 base pages. For each base page, there is a corresponding page struct. Within the HugeTLB subsystem, only the first 4 page structs are used to contain unique information about a HugeTLB page. HUGETLB_CGROUP_MIN_ORDER provides this upper limit. The only 'useful' information in the remaining page structs is the compound_head field, and this field is the same for all tail pages. By removing redundant page structs for HugeTLB pages, memory can returned to the buddy allocator for other uses. When the system boot up, every 2M HugeTLB has 512 struct page structs which size is 8 pages(sizeof(struct page) * 512 / PAGE_SIZE). HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | -------------> | 2 | | | +-----------+ +-----------+ | | | 3 | -------------> | 3 | | | +-----------+ +-----------+ | | | 4 | -------------> | 4 | | 2MB | +-----------+ +-----------+ | | | 5 | -------------> | 5 | | | +-----------+ +-----------+ | | | 6 | -------------> | 6 | | | +-----------+ +-----------+ | | | 7 | -------------> | 7 | | | +-----------+ +-----------+ | | | | | | +-----------+ The value of page->compound_head is the same for all tail pages. The first page of page structs (page 0) associated with the HugeTLB page contains the 4 page structs necessary to describe the HugeTLB. The only use of the remaining pages of page structs (page 1 to page 7) is to point to page->compound_head. Therefore, we can remap pages 2 to 7 to page 1. Only 2 pages of page structs will be used for each HugeTLB page. This will allow us to free the remaining 6 pages to the buddy allocator. Here is how things look after remapping. HugeTLB struct pages(8 pages) page frame(8 pages) +-----------+ ---virt_to_page---> +-----------+ mapping to +-----------+ | | | 0 | -------------> | 0 | | | +-----------+ +-----------+ | | | 1 | -------------> | 1 | | | +-----------+ +-----------+ | | | 2 | ----------------^ ^ ^ ^ ^ ^ | | +-----------+ | | | | | | | | 3 | ------------------+ | | | | | | +-----------+ | | | | | | | 4 | --------------------+ | | | | 2MB | +-----------+ | | | | | | 5 | ----------------------+ | | | | +-----------+ | | | | | 6 | ------------------------+ | | | +-----------+ | | | | 7 | --------------------------+ | | +-----------+ | | | | | | +-----------+ When a HugeTLB is freed to the buddy system, we should allocate 6 pages for vmemmap pages and restore the previous mapping relationship. Apart from 2MB HugeTLB page, we also have 1GB HugeTLB page. It is similar to the 2MB HugeTLB page. We also can use this approach to free the vmemmap pages. In this case, for the 1GB HugeTLB page, we can save 4094 pages. This is a very substantial gain. On our server, run some SPDK/QEMU applications which will use 1024GB HugeTLB page. With this feature enabled, we can save ~16GB (1G hugepage)/~12GB (2MB hugepage) memory. Because there are vmemmap page tables reconstruction on the freeing/allocating path, it increases some overhead. Here are some overhead analysis. 1) Allocating 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.166s user 0m0.000s sys 0m0.166s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 5476 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [16K, 32K) 4760 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ | [32K, 64K) 4 | | b) Without this patch series: # time echo 10240 > /proc/sys/vm/nr_hugepages real 0m0.067s user 0m0.000s sys 0m0.067s # bpftrace -e 'kprobe:alloc_fresh_huge_page { @start[tid] = nsecs; } kretprobe:alloc_fresh_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 10147 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 93 | | Summarize: this feature is about ~2x slower than before. 2) Freeing 10240 2MB HugeTLB pages. a) With this patch series applied: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.213s user 0m0.000s sys 0m0.213s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [8K, 16K) 6 | | [16K, 32K) 10227 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [32K, 64K) 7 | | b) Without this patch series: # time echo 0 > /proc/sys/vm/nr_hugepages real 0m0.081s user 0m0.000s sys 0m0.081s # bpftrace -e 'kprobe:free_pool_huge_page { @start[tid] = nsecs; } kretprobe:free_pool_huge_page /@start[tid]/ { @latency = hist(nsecs - @start[tid]); delete(@start[tid]); }' Attaching 2 probes... @latency: [4K, 8K) 6805 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@| [8K, 16K) 3427 |@@@@@@@@@@@@@@@@@@@@@@@@@@ | [16K, 32K) 8 | | Summary: The overhead of __free_hugepage is about ~2-3x slower than before. Although the overhead has increased, the overhead is not significant. Like Mike said, "However, remember that the majority of use cases create HugeTLB pages at or shortly after boot time and add them to the pool. So, additional overhead is at pool creation time. There is no change to 'normal run time' operations of getting a page from or returning a page to the pool (think page fault/unmap)". Despite the overhead and in addition to the memory gains from this series. The following data is obtained by Joao Martins. Very thanks to his effort. There's an additional benefit which is page (un)pinners will see an improvement and Joao presumes because there are fewer memmap pages and thus the tail/head pages are staying in cache more often. Out of the box Joao saw (when comparing linux-next against linux-next + this series) with gup_test and pinning a 16G HugeTLB file (with 1G pages): get_user_pages(): ~32k -> ~9k unpin_user_pages(): ~75k -> ~70k Usually any tight loop fetching compound_head(), or reading tail pages data (e.g. compound_head) benefit a lot. There's some unpinning inefficiencies Joao was fixing[2], but with that in added it shows even more: unpin_user_pages(): ~27k -> ~3.8k [1] https://lore.kernel.org/linux-mm/20210409205254.242291-1-mike.kravetz@oracle.com/ [2] https://lore.kernel.org/linux-mm/20210204202500.26474-1-joao.m.martins@oracle.com/ This patch (of 9): Move bootmem info registration common API to individual bootmem_info.c. And we will use {get,put}_page_bootmem() to initialize the page for the vmemmap pages or free the vmemmap pages to buddy in the later patch. So move them out of CONFIG_MEMORY_HOTPLUG_SPARSE. This is just code movement without any functional change. Link: https://lkml.kernel.org/r/20210510030027.56044-1-songmuchun@bytedance.com Link: https://lkml.kernel.org/r/20210510030027.56044-2-songmuchun@bytedance.com Signed-off-by: Muchun Song <songmuchun@bytedance.com> Acked-by: Mike Kravetz <mike.kravetz@oracle.com> Reviewed-by: Oscar Salvador <osalvador@suse.de> Reviewed-by: David Hildenbrand <david@redhat.com> Reviewed-by: Miaohe Lin <linmiaohe@huawei.com> Tested-by: Chen Huang <chenhuang5@huawei.com> Tested-by: Bodeddula Balasubramaniam <bodeddub@amazon.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Borislav Petkov <bp@alien8.de> Cc: x86@kernel.org Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Dave Hansen <dave.hansen@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Pawan Gupta <pawan.kumar.gupta@linux.intel.com> Cc: Randy Dunlap <rdunlap@infradead.org> Cc: Oliver Neukum <oneukum@suse.com> Cc: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Joerg Roedel <jroedel@suse.de> Cc: Mina Almasry <almasrymina@google.com> Cc: David Rientjes <rientjes@google.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Barry Song <song.bao.hua@hisilicon.com> Cc: HORIGUCHI NAOYA <naoya.horiguchi@nec.com> Cc: Joao Martins <joao.m.martins@oracle.com> Cc: Xiongchun Duan <duanxiongchun@bytedance.com> Cc: Balbir Singh <bsingharora@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> |
H A D | util.c | diff f5eec036 Thu Feb 22 12:46:17 MST 2024 Matthew Cassell <mcassell411@gmail.com> mm/util.c: add byte count to __vm_enough_memory failure warning Commit 44b414c8715c5dcf53288 ("mm/util.c: add warning if __vm_enough_memory fails") adds debug information which gives the process id and executable name should __vm_enough_memory() fail. Adding the number of pages to the failure message would benefit application developers and system administrators in debugging overambitious memory requests by providing a point of reference to the amount of memory causing __vm_enough_memory() to fail. 1. Set appropriate kernel tunable to reach code path for failure message: # echo 2 > /proc/sys/vm/overcommit_memory 2. Test program to generate failure - requests 1 gibibyte per iteration: #include <stdlib.h> #include <stdio.h> int main(int argc, char **argv) { for(;;) { if(malloc(1<<30) == NULL) break; printf("allocated 1 GiB\n"); } return 0; } 3. Output: Before: __vm_enough_memory: pid: 1218, comm: a.out, not enough memory for the allocation After: __vm_enough_memory: pid: 1137, comm: a.out, bytes: 1073741824, not enough memory for the allocation Link: https://lkml.kernel.org/r/20240222194617.1255-1-mcassell411@gmail.com Signed-off-by: Matthew Cassell <mcassell411@gmail.com> Cc: David Hildenbrand <david@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff c83ad36a Mon Sep 04 12:08:05 MDT 2023 Zqiang <qiang.zhang1211@gmail.com> rcu: dump vmalloc memory info safely Currently, for double invoke call_rcu(), will dump rcu_head objects memory info, if the objects is not allocated from the slab allocator, the vmalloc_dump_obj() will be invoke and the vmap_area_lock spinlock need to be held, since the call_rcu() can be invoked in interrupt context, therefore, there is a possibility of spinlock deadlock scenarios. And in Preempt-RT kernel, the rcutorture test also trigger the following lockdep warning: BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48 in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: swapper/0 preempt_count: 1, expected: 0 RCU nest depth: 1, expected: 1 3 locks held by swapper/0/1: #0: ffffffffb534ee80 (fullstop_mutex){+.+.}-{4:4}, at: torture_init_begin+0x24/0xa0 #1: ffffffffb5307940 (rcu_read_lock){....}-{1:3}, at: rcu_torture_init+0x1ec7/0x2370 #2: ffffffffb536af40 (vmap_area_lock){+.+.}-{3:3}, at: find_vmap_area+0x1f/0x70 irq event stamp: 565512 hardirqs last enabled at (565511): [<ffffffffb379b138>] __call_rcu_common+0x218/0x940 hardirqs last disabled at (565512): [<ffffffffb5804262>] rcu_torture_init+0x20b2/0x2370 softirqs last enabled at (399112): [<ffffffffb36b2586>] __local_bh_enable_ip+0x126/0x170 softirqs last disabled at (399106): [<ffffffffb43fef59>] inet_register_protosw+0x9/0x1d0 Preemption disabled at: [<ffffffffb58040c3>] rcu_torture_init+0x1f13/0x2370 CPU: 0 PID: 1 Comm: swapper/0 Tainted: G W 6.5.0-rc4-rt2-yocto-preempt-rt+ #15 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> dump_stack_lvl+0x68/0xb0 dump_stack+0x14/0x20 __might_resched+0x1aa/0x280 ? __pfx_rcu_torture_err_cb+0x10/0x10 rt_spin_lock+0x53/0x130 ? find_vmap_area+0x1f/0x70 find_vmap_area+0x1f/0x70 vmalloc_dump_obj+0x20/0x60 mem_dump_obj+0x22/0x90 __call_rcu_common+0x5bf/0x940 ? debug_smp_processor_id+0x1b/0x30 call_rcu_hurry+0x14/0x20 rcu_torture_init+0x1f82/0x2370 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_init+0x10/0x10 do_one_initcall+0x6c/0x300 ? debug_smp_processor_id+0x1b/0x30 kernel_init_freeable+0x2b9/0x540 ? __pfx_kernel_init+0x10/0x10 kernel_init+0x1f/0x150 ret_from_fork+0x40/0x50 ? __pfx_kernel_init+0x10/0x10 ret_from_fork_asm+0x1b/0x30 </TASK> The previous patch fixes this by using the deadlock-safe best-effort version of find_vm_area. However, in case of failure print the fact that the pointer was a vmalloc pointer so that we print at least something. Link: https://lkml.kernel.org/r/20230904180806.1002832-2-joel@joelfernandes.org Fixes: 98f180837a89 ("mm: Make mem_dump_obj() handle vmalloc() memory") Signed-off-by: Zqiang <qiang.zhang1211@gmail.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Reported-by: Zhen Lei <thunder.leizhen@huaweicloud.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff c83ad36a Mon Sep 04 12:08:05 MDT 2023 Zqiang <qiang.zhang1211@gmail.com> rcu: dump vmalloc memory info safely Currently, for double invoke call_rcu(), will dump rcu_head objects memory info, if the objects is not allocated from the slab allocator, the vmalloc_dump_obj() will be invoke and the vmap_area_lock spinlock need to be held, since the call_rcu() can be invoked in interrupt context, therefore, there is a possibility of spinlock deadlock scenarios. And in Preempt-RT kernel, the rcutorture test also trigger the following lockdep warning: BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48 in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: swapper/0 preempt_count: 1, expected: 0 RCU nest depth: 1, expected: 1 3 locks held by swapper/0/1: #0: ffffffffb534ee80 (fullstop_mutex){+.+.}-{4:4}, at: torture_init_begin+0x24/0xa0 #1: ffffffffb5307940 (rcu_read_lock){....}-{1:3}, at: rcu_torture_init+0x1ec7/0x2370 #2: ffffffffb536af40 (vmap_area_lock){+.+.}-{3:3}, at: find_vmap_area+0x1f/0x70 irq event stamp: 565512 hardirqs last enabled at (565511): [<ffffffffb379b138>] __call_rcu_common+0x218/0x940 hardirqs last disabled at (565512): [<ffffffffb5804262>] rcu_torture_init+0x20b2/0x2370 softirqs last enabled at (399112): [<ffffffffb36b2586>] __local_bh_enable_ip+0x126/0x170 softirqs last disabled at (399106): [<ffffffffb43fef59>] inet_register_protosw+0x9/0x1d0 Preemption disabled at: [<ffffffffb58040c3>] rcu_torture_init+0x1f13/0x2370 CPU: 0 PID: 1 Comm: swapper/0 Tainted: G W 6.5.0-rc4-rt2-yocto-preempt-rt+ #15 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> dump_stack_lvl+0x68/0xb0 dump_stack+0x14/0x20 __might_resched+0x1aa/0x280 ? __pfx_rcu_torture_err_cb+0x10/0x10 rt_spin_lock+0x53/0x130 ? find_vmap_area+0x1f/0x70 find_vmap_area+0x1f/0x70 vmalloc_dump_obj+0x20/0x60 mem_dump_obj+0x22/0x90 __call_rcu_common+0x5bf/0x940 ? debug_smp_processor_id+0x1b/0x30 call_rcu_hurry+0x14/0x20 rcu_torture_init+0x1f82/0x2370 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_init+0x10/0x10 do_one_initcall+0x6c/0x300 ? debug_smp_processor_id+0x1b/0x30 kernel_init_freeable+0x2b9/0x540 ? __pfx_kernel_init+0x10/0x10 kernel_init+0x1f/0x150 ret_from_fork+0x40/0x50 ? __pfx_kernel_init+0x10/0x10 ret_from_fork_asm+0x1b/0x30 </TASK> The previous patch fixes this by using the deadlock-safe best-effort version of find_vm_area. However, in case of failure print the fact that the pointer was a vmalloc pointer so that we print at least something. Link: https://lkml.kernel.org/r/20230904180806.1002832-2-joel@joelfernandes.org Fixes: 98f180837a89 ("mm: Make mem_dump_obj() handle vmalloc() memory") Signed-off-by: Zqiang <qiang.zhang1211@gmail.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Reported-by: Zhen Lei <thunder.leizhen@huaweicloud.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff c83ad36a Mon Sep 04 12:08:05 MDT 2023 Zqiang <qiang.zhang1211@gmail.com> rcu: dump vmalloc memory info safely Currently, for double invoke call_rcu(), will dump rcu_head objects memory info, if the objects is not allocated from the slab allocator, the vmalloc_dump_obj() will be invoke and the vmap_area_lock spinlock need to be held, since the call_rcu() can be invoked in interrupt context, therefore, there is a possibility of spinlock deadlock scenarios. And in Preempt-RT kernel, the rcutorture test also trigger the following lockdep warning: BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48 in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: swapper/0 preempt_count: 1, expected: 0 RCU nest depth: 1, expected: 1 3 locks held by swapper/0/1: #0: ffffffffb534ee80 (fullstop_mutex){+.+.}-{4:4}, at: torture_init_begin+0x24/0xa0 #1: ffffffffb5307940 (rcu_read_lock){....}-{1:3}, at: rcu_torture_init+0x1ec7/0x2370 #2: ffffffffb536af40 (vmap_area_lock){+.+.}-{3:3}, at: find_vmap_area+0x1f/0x70 irq event stamp: 565512 hardirqs last enabled at (565511): [<ffffffffb379b138>] __call_rcu_common+0x218/0x940 hardirqs last disabled at (565512): [<ffffffffb5804262>] rcu_torture_init+0x20b2/0x2370 softirqs last enabled at (399112): [<ffffffffb36b2586>] __local_bh_enable_ip+0x126/0x170 softirqs last disabled at (399106): [<ffffffffb43fef59>] inet_register_protosw+0x9/0x1d0 Preemption disabled at: [<ffffffffb58040c3>] rcu_torture_init+0x1f13/0x2370 CPU: 0 PID: 1 Comm: swapper/0 Tainted: G W 6.5.0-rc4-rt2-yocto-preempt-rt+ #15 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> dump_stack_lvl+0x68/0xb0 dump_stack+0x14/0x20 __might_resched+0x1aa/0x280 ? __pfx_rcu_torture_err_cb+0x10/0x10 rt_spin_lock+0x53/0x130 ? find_vmap_area+0x1f/0x70 find_vmap_area+0x1f/0x70 vmalloc_dump_obj+0x20/0x60 mem_dump_obj+0x22/0x90 __call_rcu_common+0x5bf/0x940 ? debug_smp_processor_id+0x1b/0x30 call_rcu_hurry+0x14/0x20 rcu_torture_init+0x1f82/0x2370 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_init+0x10/0x10 do_one_initcall+0x6c/0x300 ? debug_smp_processor_id+0x1b/0x30 kernel_init_freeable+0x2b9/0x540 ? __pfx_kernel_init+0x10/0x10 kernel_init+0x1f/0x150 ret_from_fork+0x40/0x50 ? __pfx_kernel_init+0x10/0x10 ret_from_fork_asm+0x1b/0x30 </TASK> The previous patch fixes this by using the deadlock-safe best-effort version of find_vm_area. However, in case of failure print the fact that the pointer was a vmalloc pointer so that we print at least something. Link: https://lkml.kernel.org/r/20230904180806.1002832-2-joel@joelfernandes.org Fixes: 98f180837a89 ("mm: Make mem_dump_obj() handle vmalloc() memory") Signed-off-by: Zqiang <qiang.zhang1211@gmail.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Reported-by: Zhen Lei <thunder.leizhen@huaweicloud.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff c83ad36a Mon Sep 04 12:08:05 MDT 2023 Zqiang <qiang.zhang1211@gmail.com> rcu: dump vmalloc memory info safely Currently, for double invoke call_rcu(), will dump rcu_head objects memory info, if the objects is not allocated from the slab allocator, the vmalloc_dump_obj() will be invoke and the vmap_area_lock spinlock need to be held, since the call_rcu() can be invoked in interrupt context, therefore, there is a possibility of spinlock deadlock scenarios. And in Preempt-RT kernel, the rcutorture test also trigger the following lockdep warning: BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48 in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 1, name: swapper/0 preempt_count: 1, expected: 0 RCU nest depth: 1, expected: 1 3 locks held by swapper/0/1: #0: ffffffffb534ee80 (fullstop_mutex){+.+.}-{4:4}, at: torture_init_begin+0x24/0xa0 #1: ffffffffb5307940 (rcu_read_lock){....}-{1:3}, at: rcu_torture_init+0x1ec7/0x2370 #2: ffffffffb536af40 (vmap_area_lock){+.+.}-{3:3}, at: find_vmap_area+0x1f/0x70 irq event stamp: 565512 hardirqs last enabled at (565511): [<ffffffffb379b138>] __call_rcu_common+0x218/0x940 hardirqs last disabled at (565512): [<ffffffffb5804262>] rcu_torture_init+0x20b2/0x2370 softirqs last enabled at (399112): [<ffffffffb36b2586>] __local_bh_enable_ip+0x126/0x170 softirqs last disabled at (399106): [<ffffffffb43fef59>] inet_register_protosw+0x9/0x1d0 Preemption disabled at: [<ffffffffb58040c3>] rcu_torture_init+0x1f13/0x2370 CPU: 0 PID: 1 Comm: swapper/0 Tainted: G W 6.5.0-rc4-rt2-yocto-preempt-rt+ #15 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.16.2-0-gea1b7a073390-prebuilt.qemu.org 04/01/2014 Call Trace: <TASK> dump_stack_lvl+0x68/0xb0 dump_stack+0x14/0x20 __might_resched+0x1aa/0x280 ? __pfx_rcu_torture_err_cb+0x10/0x10 rt_spin_lock+0x53/0x130 ? find_vmap_area+0x1f/0x70 find_vmap_area+0x1f/0x70 vmalloc_dump_obj+0x20/0x60 mem_dump_obj+0x22/0x90 __call_rcu_common+0x5bf/0x940 ? debug_smp_processor_id+0x1b/0x30 call_rcu_hurry+0x14/0x20 rcu_torture_init+0x1f82/0x2370 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_leak_cb+0x10/0x10 ? __pfx_rcu_torture_init+0x10/0x10 do_one_initcall+0x6c/0x300 ? debug_smp_processor_id+0x1b/0x30 kernel_init_freeable+0x2b9/0x540 ? __pfx_kernel_init+0x10/0x10 kernel_init+0x1f/0x150 ret_from_fork+0x40/0x50 ? __pfx_kernel_init+0x10/0x10 ret_from_fork_asm+0x1b/0x30 </TASK> The previous patch fixes this by using the deadlock-safe best-effort version of find_vm_area. However, in case of failure print the fact that the pointer was a vmalloc pointer so that we print at least something. Link: https://lkml.kernel.org/r/20230904180806.1002832-2-joel@joelfernandes.org Fixes: 98f180837a89 ("mm: Make mem_dump_obj() handle vmalloc() memory") Signed-off-by: Zqiang <qiang.zhang1211@gmail.com> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org> Reported-by: Zhen Lei <thunder.leizhen@huaweicloud.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 3d2c9087 Mon Aug 21 10:08:48 MDT 2023 David Hildenbrand <david@redhat.com> mm/swap: inline folio_set_swap_entry() and folio_swap_entry() Let's simply work on the folio directly and remove the helpers. Link: https://lkml.kernel.org/r/20230821160849.531668-4-david@redhat.com Signed-off-by: David Hildenbrand <david@redhat.com> Suggested-by: Matthew Wilcox <willy@infradead.org> Reviewed-by: Chris Li <chrisl@kernel.org> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Hugh Dickins <hughd@google.com> Cc: Peter Xu <peterx@redhat.com> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Cc: Will Deacon <will@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff fc1878ec Fri Jun 30 21:28:53 MDT 2023 ZhangPeng <zhangpeng362@huawei.com> mm: remove page_rmapping() After converting the last user to folio_raw_mapping(), we can safely remove the function. Link: https://lkml.kernel.org/r/20230701032853.258697-3-zhangpeng362@huawei.com Signed-off-by: ZhangPeng <zhangpeng362@huawei.com> Reviewed-by: Sidhartha Kumar <sidhartha.kumar@oracle.com> Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Kefeng Wang <wangkefeng.wang@huawei.com> Cc: Nanyong Sun <sunnanyong@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 9becb688 Fri Apr 22 12:41:38 MDT 2022 Linus Torvalds <torvalds@linux-foundation.org> kvmalloc: use vmalloc_huge for vmalloc allocations Since commit 559089e0a93d ("vmalloc: replace VM_NO_HUGE_VMAP with VM_ALLOW_HUGE_VMAP"), the use of hugepage mappings for vmalloc is an opt-in strategy, because it caused a number of problems that weren't noticed until x86 enabled it too. One of the issues was fixed by Nick Piggin in commit 3b8000ae185c ("mm/vmalloc: huge vmalloc backing pages should be split rather than compound"), but I'm still worried about page protection issues, and VM_FLUSH_RESET_PERMS in particular. However, like the hash table allocation case (commit f2edd118d02d: "page_alloc: use vmalloc_huge for large system hash"), the use of kvmalloc() should be safe from any such games, since the returned pointer might be a SLUB allocation, and as such no user should reasonably be using it in any odd ways. We also know that the allocations are fairly large, since it falls back to the vmalloc case only when a kmalloc() fails. So using a hugepage mapping seems both safe and relevant. This patch does show a weakness in the opt-in strategy: since the opt-in flag is in the 'vm_flags', not the usual gfp_t allocation flags, very few of the usual interfaces actually expose it. That's not much of an issue in this case that already used one of the fairly specialized low-level vmalloc interfaces for the allocation, but for a lot of other vmalloc() users that might want to opt in, it's going to be very inconvenient. We'll either have to fix any compatibility problems, or expose it in the gfp flags (__GFP_COMP would have made a lot of sense) to allow normal vmalloc() users to use hugepage mappings. That said, the cases that really matter were probably already taken care of by the hash tabel allocation. Link: https://lore.kernel.org/all/20220415164413.2727220-1-song@kernel.org/ Link: https://lore.kernel.org/all/CAHk-=whao=iosX1s5Z4SF-ZGa-ebAukJoAdUJFk5SPwnofV+Vg@mail.gmail.com/ Cc: Nicholas Piggin <npiggin@gmail.com> Cc: Paul Menzel <pmenzel@molgen.mpg.de> Cc: Song Liu <songliubraving@fb.com> Cc: Rick Edgecombe <rick.p.edgecombe@intel.com> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 56f3547b Fri Aug 07 00:23:15 MDT 2020 Feng Tang <feng.tang@intel.com> mm: adjust vm_committed_as_batch according to vm overcommit policy When checking a performance change for will-it-scale scalability mmap test [1], we found very high lock contention for spinlock of percpu counter 'vm_committed_as': 94.14% 0.35% [kernel.kallsyms] [k] _raw_spin_lock_irqsave 48.21% _raw_spin_lock_irqsave;percpu_counter_add_batch;__vm_enough_memory;mmap_region;do_mmap; 45.91% _raw_spin_lock_irqsave;percpu_counter_add_batch;__do_munmap; Actually this heavy lock contention is not always necessary. The 'vm_committed_as' needs to be very precise when the strict OVERCOMMIT_NEVER policy is set, which requires a rather small batch number for the percpu counter. So keep 'batch' number unchanged for strict OVERCOMMIT_NEVER policy, and lift it to 64X for OVERCOMMIT_ALWAYS and OVERCOMMIT_GUESS policies. Also add a sysctl handler to adjust it when the policy is reconfigured. Benchmark with the same testcase in [1] shows 53% improvement on a 8C/16T desktop, and 2097%(20X) on a 4S/72C/144T server. We tested with test platforms in 0day (server, desktop and laptop), and 80%+ platforms shows improvements with that test. And whether it shows improvements depends on if the test mmap size is bigger than the batch number computed. And if the lift is 16X, 1/3 of the platforms will show improvements, though it should help the mmap/unmap usage generally, as Michal Hocko mentioned: : I believe that there are non-synthetic worklaods which would benefit from : a larger batch. E.g. large in memory databases which do large mmaps : during startups from multiple threads. [1] https://lore.kernel.org/lkml/20200305062138.GI5972@shao2-debian/ Signed-off-by: Feng Tang <feng.tang@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Michal Hocko <mhocko@suse.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Qian Cai <cai@lca.pw> Cc: Kees Cook <keescook@chromium.org> Cc: Andi Kleen <andi.kleen@intel.com> Cc: Tim Chen <tim.c.chen@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Huang Ying <ying.huang@intel.com> Cc: Christoph Lameter <cl@linux.com> Cc: Dennis Zhou <dennis@kernel.org> Cc: Haiyang Zhang <haiyangz@microsoft.com> Cc: kernel test robot <rong.a.chen@intel.com> Cc: "K. Y. Srinivasan" <kys@microsoft.com> Cc: Tejun Heo <tj@kernel.org> Link: http://lkml.kernel.org/r/1589611660-89854-4-git-send-email-feng.tang@intel.com Link: http://lkml.kernel.org/r/1592725000-73486-4-git-send-email-feng.tang@intel.com Link: http://lkml.kernel.org/r/1594389708-60781-5-git-send-email-feng.tang@intel.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 4e2ee51e Fri Aug 07 00:23:07 MDT 2020 Feng Tang <feng.tang@intel.com> mm/util.c: make vm_memory_committed() more accurate percpu_counter_sum_positive() will provide more accurate info. As with percpu_counter_read_positive(), in worst case the deviation could be 'batch * nr_cpus', which is totalram_pages/256 for now, and will be more when the batch gets enlarged. Its time cost is about 800 nanoseconds on a 2C/4T platform and 2~3 microseconds on a 2S/36C/72T Skylake server in normal case, and in worst case where vm_committed_as's spinlock is under severe contention, it costs 30~40 microseconds for the 2S/36C/72T Skylake sever, which should be fine for its only two users: /proc/meminfo and HyperV balloon driver's status trace per second. Signed-off-by: Feng Tang <feng.tang@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Michal Hocko <mhocko@suse.com> # for /proc/meminfo Cc: "K. Y. Srinivasan" <kys@microsoft.com> Cc: Haiyang Zhang <haiyangz@microsoft.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Qian Cai <cai@lca.pw> Cc: Andi Kleen <andi.kleen@intel.com> Cc: Tim Chen <tim.c.chen@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Huang Ying <ying.huang@intel.com> Cc: Christoph Lameter <cl@linux.com> Cc: Dennis Zhou <dennis@kernel.org> Cc: Kees Cook <keescook@chromium.org> Cc: kernel test robot <rong.a.chen@intel.com> Cc: Tejun Heo <tj@kernel.org> Link: http://lkml.kernel.org/r/1592725000-73486-3-git-send-email-feng.tang@intel.com Link: http://lkml.kernel.org/r/1594389708-60781-3-git-send-email-feng.tang@intel.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 4e2ee51e Fri Aug 07 00:23:07 MDT 2020 Feng Tang <feng.tang@intel.com> mm/util.c: make vm_memory_committed() more accurate percpu_counter_sum_positive() will provide more accurate info. As with percpu_counter_read_positive(), in worst case the deviation could be 'batch * nr_cpus', which is totalram_pages/256 for now, and will be more when the batch gets enlarged. Its time cost is about 800 nanoseconds on a 2C/4T platform and 2~3 microseconds on a 2S/36C/72T Skylake server in normal case, and in worst case where vm_committed_as's spinlock is under severe contention, it costs 30~40 microseconds for the 2S/36C/72T Skylake sever, which should be fine for its only two users: /proc/meminfo and HyperV balloon driver's status trace per second. Signed-off-by: Feng Tang <feng.tang@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Michal Hocko <mhocko@suse.com> # for /proc/meminfo Cc: "K. Y. Srinivasan" <kys@microsoft.com> Cc: Haiyang Zhang <haiyangz@microsoft.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Qian Cai <cai@lca.pw> Cc: Andi Kleen <andi.kleen@intel.com> Cc: Tim Chen <tim.c.chen@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Huang Ying <ying.huang@intel.com> Cc: Christoph Lameter <cl@linux.com> Cc: Dennis Zhou <dennis@kernel.org> Cc: Kees Cook <keescook@chromium.org> Cc: kernel test robot <rong.a.chen@intel.com> Cc: Tejun Heo <tj@kernel.org> Link: http://lkml.kernel.org/r/1592725000-73486-3-git-send-email-feng.tang@intel.com Link: http://lkml.kernel.org/r/1594389708-60781-3-git-send-email-feng.tang@intel.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> diff 4e2ee51e Fri Aug 07 00:23:07 MDT 2020 Feng Tang <feng.tang@intel.com> mm/util.c: make vm_memory_committed() more accurate percpu_counter_sum_positive() will provide more accurate info. As with percpu_counter_read_positive(), in worst case the deviation could be 'batch * nr_cpus', which is totalram_pages/256 for now, and will be more when the batch gets enlarged. Its time cost is about 800 nanoseconds on a 2C/4T platform and 2~3 microseconds on a 2S/36C/72T Skylake server in normal case, and in worst case where vm_committed_as's spinlock is under severe contention, it costs 30~40 microseconds for the 2S/36C/72T Skylake sever, which should be fine for its only two users: /proc/meminfo and HyperV balloon driver's status trace per second. Signed-off-by: Feng Tang <feng.tang@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Acked-by: Michal Hocko <mhocko@suse.com> # for /proc/meminfo Cc: "K. Y. Srinivasan" <kys@microsoft.com> Cc: Haiyang Zhang <haiyangz@microsoft.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Qian Cai <cai@lca.pw> Cc: Andi Kleen <andi.kleen@intel.com> Cc: Tim Chen <tim.c.chen@intel.com> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Huang Ying <ying.huang@intel.com> Cc: Christoph Lameter <cl@linux.com> Cc: Dennis Zhou <dennis@kernel.org> Cc: Kees Cook <keescook@chromium.org> Cc: kernel test robot <rong.a.chen@intel.com> Cc: Tejun Heo <tj@kernel.org> Link: http://lkml.kernel.org/r/1592725000-73486-3-git-send-email-feng.tang@intel.com Link: http://lkml.kernel.org/r/1594389708-60781-3-git-send-email-feng.tang@intel.com Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> |
H A D | vmalloc.c | diff d0936029 Tue Jan 02 11:46:26 MST 2024 Uladzislau Rezki (Sony) <urezki@gmail.com> mm: vmalloc: remove global vmap_area_root rb-tree Store allocated objects in a separate nodes. A va->va_start address is converted into a correct node where it should be placed and resided. An addr_to_node() function is used to do a proper address conversion to determine a node that contains a VA. Such approach balances VAs across nodes as a result an access becomes scalable. Number of nodes in a system depends on number of CPUs. Please note: 1. As of now allocated VAs are bound to a node-0. It means the patch does not give any difference comparing with a current behavior; 2. The global vmap_area_lock, vmap_area_root are removed as there is no need in it anymore. The vmap_area_list is still kept and is _empty_. It is exported for a kexec only; 3. The vmallocinfo and vread() have to be reworked to be able to handle multiple nodes. [urezki@gmail.com: mark vmap_init_free_space() with __init tag] Link: https://lkml.kernel.org/r/20240111132628.299644-1-urezki@gmail.com [urezki@gmail.com: fix a wrong value passed to __find_vmap_area()] Link: https://lkml.kernel.org/r/20240111121104.180993-1-urezki@gmail.com Link: https://lkml.kernel.org/r/20240102184633.748113-5-urezki@gmail.com Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com> Reviewed-by: Baoquan He <bhe@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Lorenzo Stoakes <lstoakes@gmail.com> Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Joel Fernandes (Google) <joel@joelfernandes.org> Cc: Kazuhito Hagio <k-hagio-ab@nec.com> Cc: Liam R. Howlett <Liam.Howlett@oracle.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com> Cc: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 5b75b8e1 Tue Jan 02 11:46:24 MST 2024 Uladzislau Rezki (Sony) <urezki@gmail.com> mm: vmalloc: rename adjust_va_to_fit_type() function This patch renames the adjust_va_to_fit_type() function to va_clip() which is shorter and more expressive. There is no a functional change as a result of this patch. Link: https://lkml.kernel.org/r/20240102184633.748113-3-urezki@gmail.com Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com> Reviewed-by: Baoquan He <bhe@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Joel Fernandes (Google) <joel@joelfernandes.org> Cc: Kazuhito Hagio <k-hagio-ab@nec.com> Cc: Liam R. Howlett <Liam.Howlett@oracle.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com> Cc: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 38f6b9af Tue Jan 02 11:46:23 MST 2024 Uladzislau Rezki (Sony) <urezki@gmail.com> mm: vmalloc: add va_alloc() helper Patch series "Mitigate a vmap lock contention", v3. 1. Motivation - Offload global vmap locks making it scaled to number of CPUS; - If possible and there is an agreement, we can remove the "Per cpu kva allocator" to make the vmap code to be more simple; - There were complaints from XFS folk that a vmalloc might be contented on their workloads. 2. Design(high level overview) We introduce an effective vmap node logic. A node behaves as independent entity to serve an allocation request directly(if possible) from its pool. That way it bypasses a global vmap space that is protected by its own lock. An access to pools are serialized by CPUs. Number of nodes are equal to number of CPUs in a system. Please note the high threshold is bound to 128 nodes. Pools are size segregated and populated based on system demand. The maximum alloc request that can be stored into a segregated storage is 256 pages. The lazily drain path decays a pool by 25% as a first step and as second populates it by fresh freed VAs for reuse instead of returning them into a global space. When a VA is obtained(alloc path), it is stored in separate nodes. A va->va_start address is converted into a correct node where it should be placed and resided. Doing so we balance VAs across the nodes as a result an access becomes scalable. The addr_to_node() function does a proper address conversion to a correct node. A vmap space is divided on segments with fixed size, it is 16 pages. That way any address can be associated with a segment number. Number of segments are equal to num_possible_cpus() but not grater then 128. The numeration starts from 0. See below how it is converted: static inline unsigned int addr_to_node_id(unsigned long addr) { return (addr / zone_size) % nr_nodes; } On a free path, a VA can be easily found by converting its "va_start" address to a certain node it resides. It is moved from "busy" data to "lazy" data structure. Later on, as noted earlier, the lazy kworker decays each node pool and populates it by fresh incoming VAs. Please note, a VA is returned to a node that did an alloc request. 3. Test on AMD Ryzen Threadripper 3970X 32-Core Processor sudo ./test_vmalloc.sh run_test_mask=7 nr_threads=64 <default perf> 94.41% 0.89% [kernel] [k] _raw_spin_lock 93.35% 93.07% [kernel] [k] native_queued_spin_lock_slowpath 76.13% 0.28% [kernel] [k] __vmalloc_node_range 72.96% 0.81% [kernel] [k] alloc_vmap_area 56.94% 0.00% [kernel] [k] __get_vm_area_node 41.95% 0.00% [kernel] [k] vmalloc 37.15% 0.01% [test_vmalloc] [k] full_fit_alloc_test 35.17% 0.00% [kernel] [k] ret_from_fork_asm 35.17% 0.00% [kernel] [k] ret_from_fork 35.17% 0.00% [kernel] [k] kthread 35.08% 0.00% [test_vmalloc] [k] test_func 34.45% 0.00% [test_vmalloc] [k] fix_size_alloc_test 28.09% 0.01% [test_vmalloc] [k] long_busy_list_alloc_test 23.53% 0.25% [kernel] [k] vfree.part.0 21.72% 0.00% [kernel] [k] remove_vm_area 20.08% 0.21% [kernel] [k] find_unlink_vmap_area 2.34% 0.61% [kernel] [k] free_vmap_area_noflush <default perf> vs <patch-series perf> 82.32% 0.22% [test_vmalloc] [k] long_busy_list_alloc_test 63.36% 0.02% [kernel] [k] vmalloc 63.34% 2.64% [kernel] [k] __vmalloc_node_range 30.42% 4.46% [kernel] [k] vfree.part.0 28.98% 2.51% [kernel] [k] __alloc_pages_bulk 27.28% 0.19% [kernel] [k] __get_vm_area_node 26.13% 1.50% [kernel] [k] alloc_vmap_area 21.72% 21.67% [kernel] [k] clear_page_rep 19.51% 2.43% [kernel] [k] _raw_spin_lock 16.61% 16.51% [kernel] [k] native_queued_spin_lock_slowpath 13.40% 2.07% [kernel] [k] free_unref_page 10.62% 0.01% [kernel] [k] remove_vm_area 9.02% 8.73% [kernel] [k] insert_vmap_area 8.94% 0.00% [kernel] [k] ret_from_fork_asm 8.94% 0.00% [kernel] [k] ret_from_fork 8.94% 0.00% [kernel] [k] kthread 8.29% 0.00% [test_vmalloc] [k] test_func 7.81% 0.05% [test_vmalloc] [k] full_fit_alloc_test 5.30% 4.73% [kernel] [k] purge_vmap_node 4.47% 2.65% [kernel] [k] free_vmap_area_noflush <patch-series perf> confirms that a native_queued_spin_lock_slowpath goes down to 16.51% percent from 93.07%. The throughput is ~12x higher: urezki@pc638:~$ time sudo ./test_vmalloc.sh run_test_mask=7 nr_threads=64 Run the test with following parameters: run_test_mask=7 nr_threads=64 Done. Check the kernel ring buffer to see the summary. real 10m51.271s user 0m0.013s sys 0m0.187s urezki@pc638:~$ urezki@pc638:~$ time sudo ./test_vmalloc.sh run_test_mask=7 nr_threads=64 Run the test with following parameters: run_test_mask=7 nr_threads=64 Done. Check the kernel ring buffer to see the summary. real 0m51.301s user 0m0.015s sys 0m0.040s urezki@pc638:~$ This patch (of 11): Currently __alloc_vmap_area() function contains an open codded logic that finds and adjusts a VA based on allocation request. Introduce a va_alloc() helper that adjusts found VA only. There is no a functional change as a result of this patch. Link: https://lkml.kernel.org/r/20240102184633.748113-1-urezki@gmail.com Link: https://lkml.kernel.org/r/20240102184633.748113-2-urezki@gmail.com Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com> Reviewed-by: Baoquan He <bhe@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Joel Fernandes (Google) <joel@joelfernandes.org> Cc: Liam R. Howlett <Liam.Howlett@oracle.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sony.com> Cc: Paul E. McKenney <paulmck@kernel.org> Cc: Kazuhito Hagio <k-hagio-ab@nec.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff d7bca919 Fri Mar 08 10:12:54 MST 2024 Alexei Starovoitov <ast@kernel.org> mm: Introduce vmap_page_range() to map pages in PCI address space ioremap_page_range() should be used for ranges within vmalloc range only. The vmalloc ranges are allocated by get_vm_area(). PCI has "resource" allocator that manages PCI_IOBASE, IO_SPACE_LIMIT address range, hence introduce vmap_page_range() to be used exclusively to map pages in PCI address space. Fixes: 3e49a866c9dc ("mm: Enforce VM_IOREMAP flag and range in ioremap_page_range.") Reported-by: Miguel Ojeda <ojeda@kernel.org> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net> Reviewed-by: Christoph Hellwig <hch@lst.de> Tested-by: Miguel Ojeda <ojeda@kernel.org> Link: https://lore.kernel.org/bpf/CANiq72ka4rir+RTN2FQoT=Vvprp_Ao-CvoYEkSNqtSY+RZj+AA@mail.gmail.com diff e6f798225 Mon Mar 04 20:05:16 MST 2024 Alexei Starovoitov <ast@kernel.org> mm: Introduce VM_SPARSE kind and vm_area_[un]map_pages(). vmap/vmalloc APIs are used to map a set of pages into contiguous kernel virtual space. get_vm_area() with appropriate flag is used to request an area of kernel address range. It's used for vmalloc, vmap, ioremap, xen use cases. - vmalloc use case dominates the usage. Such vm areas have VM_ALLOC flag. - the areas created by vmap() function should be tagged with VM_MAP. - ioremap areas are tagged with VM_IOREMAP. BPF would like to extend the vmap API to implement a lazily-populated sparse, yet contiguous kernel virtual space. Introduce VM_SPARSE flag and vm_area_map_pages(area, start_addr, count, pages) API to map a set of pages within a given area. It has the same sanity checks as vmap() does. It also checks that get_vm_area() was created with VM_SPARSE flag which identifies such areas in /proc/vmallocinfo and returns zero pages on read through /proc/kcore. The next commits will introduce bpf_arena which is a sparsely populated shared memory region between bpf program and user space process. It will map privately-managed pages into a sparse vm area with the following steps: // request virtual memory region during bpf prog verification area = get_vm_area(area_size, VM_SPARSE); // on demand vm_area_map_pages(area, kaddr, kend, pages); vm_area_unmap_pages(area, kaddr, kend); // after bpf program is detached and unloaded free_vm_area(area); Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Pasha Tatashin <pasha.tatashin@soleen.com> Link: https://lore.kernel.org/bpf/20240305030516.41519-3-alexei.starovoitov@gmail.com diff 3e49a866 Mon Mar 04 20:05:15 MST 2024 Alexei Starovoitov <ast@kernel.org> mm: Enforce VM_IOREMAP flag and range in ioremap_page_range. There are various users of get_vm_area() + ioremap_page_range() APIs. Enforce that get_vm_area() was requested as VM_IOREMAP type and range passed to ioremap_page_range() matches created vm_area to avoid accidentally ioremap-ing into wrong address range. Signed-off-by: Alexei Starovoitov <ast@kernel.org> Signed-off-by: Andrii Nakryiko <andrii@kernel.org> Reviewed-by: Christoph Hellwig <hch@lst.de> Link: https://lore.kernel.org/bpf/20240305030516.41519-2-alexei.starovoitov@gmail.com diff a50420c7 Wed Aug 09 10:46:33 MDT 2023 Alexandre Ghiti <alexghiti@rivosinc.com> mm: add a call to flush_cache_vmap() in vmap_pfn() flush_cache_vmap() must be called after new vmalloc mappings are installed in the page table in order to allow architectures to make sure the new mapping is visible. It could lead to a panic since on some architectures (like powerpc), the page table walker could see the wrong pte value and trigger a spurious page fault that can not be resolved (see commit f1cb8f9beba8 ("powerpc/64s/radix: avoid ptesync after set_pte and ptep_set_access_flags")). But actually the patch is aiming at riscv: the riscv specification allows the caching of invalid entries in the TLB, and since we recently removed the vmalloc page fault handling, we now need to emit a tlb shootdown whenever a new vmalloc mapping is emitted (https://lore.kernel.org/linux-riscv/20230725132246.817726-1-alexghiti@rivosinc.com/). That's a temporary solution, there are ways to avoid that :) Link: https://lkml.kernel.org/r/20230809164633.1556126-1-alexghiti@rivosinc.com Fixes: 3e9a9e256b1e ("mm: add a vmap_pfn function") Reported-by: Dylan Jhong <dylan@andestech.com> Closes: https://lore.kernel.org/linux-riscv/ZMytNY2J8iyjbPPy@atctrx.andestech.com/ Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Palmer Dabbelt <palmer@rivosinc.com> Acked-by: Palmer Dabbelt <palmer@rivosinc.com> Reviewed-by: Dylan Jhong <dylan@andestech.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff b3f78e74 Fri Jun 02 03:29:46 MDT 2023 Ryan Roberts <ryan.roberts@arm.com> mm: vmalloc must set pte via arch code Patch series "Fixes for pte encapsulation bypasses", v3. A series to improve the encapsulation of pte entries by disallowing non-arch code from directly dereferencing pte_t pointers. This patch (of 4): It is bad practice to directly set pte entries within a pte table. Instead all modifications must go through arch-provided helpers such as set_pte_at() to give the arch code visibility and allow it to check (and potentially modify) the operation. Link: https://lkml.kernel.org/r/20230602092949.545577-1-ryan.roberts@arm.com Link: https://lkml.kernel.org/r/20230602092949.545577-2-ryan.roberts@arm.com Fixes: 3e9a9e256b1e ("mm: add a vmap_pfn function") Signed-off-by: Ryan Roberts <ryan.roberts@arm.com> Reviewed-by: Zi Yan <ziy@nvidia.com> Acked-by: Lorenzo Stoakes <lstoakes@gmail.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com> Reviewed-by: Mike Rapoport (IBM) <rppt@kernel.org> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: SeongJae Park <sj@kernel.org> Cc: Yu Zhao <yuzhao@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 77e50af0 Thu May 25 06:57:09 MDT 2023 Thomas Gleixner <tglx@linutronix.de> mm/vmalloc: dont purge usable blocks unnecessarily Purging fragmented blocks is done unconditionally in several contexts: 1) From drain_vmap_area_work(), when the number of lazy to be freed vmap_areas reached the threshold 2) Reclaiming vmalloc address space from pcpu_get_vm_areas() 3) _vm_unmap_aliases() #1 There is no reason to zap fragmented vmap blocks unconditionally, simply because reclaiming all lazy areas drains at least 32MB * fls(num_online_cpus()) per invocation which is plenty. #2 Reclaiming when running out of space or due to memory pressure makes a lot of sense #3 _unmap_aliases() requires to touch everything because the caller has no clue which vmap_area used a particular page last and the vmap_area lost that information too. Except for the vfree + VM_FLUSH_RESET_PERMS case, which removes the vmap area first and then cares about the flush. That in turn requires a full walk of _all_ vmap areas including the one which was just added to the purge list. But as this has to be flushed anyway this is an opportunity to combine outstanding TLB flushes and do the housekeeping of purging freed areas, but like #1 there is no real good reason to zap usable vmap blocks unconditionally. Add a @force_purge argument to the newly split out block purge function and if not true only purge fragmented blocks which have less than 1/4 of their capacity left. Rename purge_vmap_area_lazy() to reclaim_and_purge_vmap_areas() to make it clear what the function does. [lstoakes@gmail.com: correct VMAP_PURGE_THRESHOLD check] Link: https://lkml.kernel.org/r/3e92ef61-b910-4576-88e7-cf43211fd4e7@lucifer.local Link: https://lkml.kernel.org/r/20230525124504.864005691@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com> Reviewed-by: Baoquan He <bhe@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 77e50af0 Thu May 25 06:57:09 MDT 2023 Thomas Gleixner <tglx@linutronix.de> mm/vmalloc: dont purge usable blocks unnecessarily Purging fragmented blocks is done unconditionally in several contexts: 1) From drain_vmap_area_work(), when the number of lazy to be freed vmap_areas reached the threshold 2) Reclaiming vmalloc address space from pcpu_get_vm_areas() 3) _vm_unmap_aliases() #1 There is no reason to zap fragmented vmap blocks unconditionally, simply because reclaiming all lazy areas drains at least 32MB * fls(num_online_cpus()) per invocation which is plenty. #2 Reclaiming when running out of space or due to memory pressure makes a lot of sense #3 _unmap_aliases() requires to touch everything because the caller has no clue which vmap_area used a particular page last and the vmap_area lost that information too. Except for the vfree + VM_FLUSH_RESET_PERMS case, which removes the vmap area first and then cares about the flush. That in turn requires a full walk of _all_ vmap areas including the one which was just added to the purge list. But as this has to be flushed anyway this is an opportunity to combine outstanding TLB flushes and do the housekeeping of purging freed areas, but like #1 there is no real good reason to zap usable vmap blocks unconditionally. Add a @force_purge argument to the newly split out block purge function and if not true only purge fragmented blocks which have less than 1/4 of their capacity left. Rename purge_vmap_area_lazy() to reclaim_and_purge_vmap_areas() to make it clear what the function does. [lstoakes@gmail.com: correct VMAP_PURGE_THRESHOLD check] Link: https://lkml.kernel.org/r/3e92ef61-b910-4576-88e7-cf43211fd4e7@lucifer.local Link: https://lkml.kernel.org/r/20230525124504.864005691@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com> Reviewed-by: Baoquan He <bhe@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 77e50af0 Thu May 25 06:57:09 MDT 2023 Thomas Gleixner <tglx@linutronix.de> mm/vmalloc: dont purge usable blocks unnecessarily Purging fragmented blocks is done unconditionally in several contexts: 1) From drain_vmap_area_work(), when the number of lazy to be freed vmap_areas reached the threshold 2) Reclaiming vmalloc address space from pcpu_get_vm_areas() 3) _vm_unmap_aliases() #1 There is no reason to zap fragmented vmap blocks unconditionally, simply because reclaiming all lazy areas drains at least 32MB * fls(num_online_cpus()) per invocation which is plenty. #2 Reclaiming when running out of space or due to memory pressure makes a lot of sense #3 _unmap_aliases() requires to touch everything because the caller has no clue which vmap_area used a particular page last and the vmap_area lost that information too. Except for the vfree + VM_FLUSH_RESET_PERMS case, which removes the vmap area first and then cares about the flush. That in turn requires a full walk of _all_ vmap areas including the one which was just added to the purge list. But as this has to be flushed anyway this is an opportunity to combine outstanding TLB flushes and do the housekeeping of purging freed areas, but like #1 there is no real good reason to zap usable vmap blocks unconditionally. Add a @force_purge argument to the newly split out block purge function and if not true only purge fragmented blocks which have less than 1/4 of their capacity left. Rename purge_vmap_area_lazy() to reclaim_and_purge_vmap_areas() to make it clear what the function does. [lstoakes@gmail.com: correct VMAP_PURGE_THRESHOLD check] Link: https://lkml.kernel.org/r/3e92ef61-b910-4576-88e7-cf43211fd4e7@lucifer.local Link: https://lkml.kernel.org/r/20230525124504.864005691@linutronix.de Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com> Reviewed-by: Baoquan He <bhe@redhat.com> Cc: Christoph Hellwig <hch@lst.de> Cc: Lorenzo Stoakes <lstoakes@gmail.com> Cc: Uladzislau Rezki (Sony) <urezki@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> |
Completed in 691 milliseconds