Searched +hist:2 +hist:f5a55c5 (Results 1 - 5 of 5) sorted by relevance
/linux-master/drivers/i2c/ | ||
H A D | i2c-core-of.c | diff 2f5a55c5 Sat May 02 06:18:35 MDT 2020 Wolfram Sang <wsa@kernel.org> i2c: use my kernel.org address from now on The old email is still active, but for easier handling, I am going to use my kernel.org address from now on. Also, add a mailmap for the now defunct Pengutronix address. Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2f5a55c5 Sat May 02 06:18:35 MDT 2020 Wolfram Sang <wsa@kernel.org> i2c: use my kernel.org address from now on The old email is still active, but for easier handling, I am going to use my kernel.org address from now on. Also, add a mailmap for the now defunct Pengutronix address. Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2874c5fd Mon May 27 00:55:01 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 152 Based on 1 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 3029 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Allison Randal <allison@lohutok.net> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070032.746973796@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> |
H A D | i2c-core-base.c | diff 6af79f7f Fri Sep 29 03:19:52 MDT 2023 Wolfram Sang <wsa+renesas@sang-engineering.com> i2c: fix memleak in i2c_new_client_device() Yang Yingliang reported a memleak: === I got memory leak as follows when doing fault injection test: unreferenced object 0xffff888014aec078 (size 8): comm "xrun", pid 356, jiffies 4294910619 (age 16.332s) hex dump (first 8 bytes): 31 2d 30 30 31 63 00 00 1-001c.. backtrace: [<00000000eb56c0a9>] __kmalloc_track_caller+0x1a6/0x300 [<000000000b220ea3>] kvasprintf+0xad/0x140 [<00000000b83203e5>] kvasprintf_const+0x62/0x190 [<000000002a5eab37>] kobject_set_name_vargs+0x56/0x140 [<00000000300ac279>] dev_set_name+0xb0/0xe0 [<00000000b66ebd6f>] i2c_new_client_device+0x7e4/0x9a0 If device_register() returns error in i2c_new_client_device(), the name allocated by i2c_dev_set_name() need be freed. As comment of device_register() says, it should use put_device() to give up the reference in the error path. === I think this solution is less intrusive and more robust than he originally proposed solutions, though. Reported-by: Yang Yingliang <yangyingliang@huawei.com> Closes: http://patchwork.ozlabs.org/project/linux-i2c/patch/20221124085448.3620240-1-yangyingliang@huawei.com/ Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2f5a55c5 Sat May 02 06:18:35 MDT 2020 Wolfram Sang <wsa@kernel.org> i2c: use my kernel.org address from now on The old email is still active, but for easier handling, I am going to use my kernel.org address from now on. Also, add a mailmap for the now defunct Pengutronix address. Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2f5a55c5 Sat May 02 06:18:35 MDT 2020 Wolfram Sang <wsa@kernel.org> i2c: use my kernel.org address from now on The old email is still active, but for easier handling, I am going to use my kernel.org address from now on. Also, add a mailmap for the now defunct Pengutronix address. Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2c2f00ab Thu Dec 05 17:23:22 MST 2019 Wolfram Sang <wsa+renesas@sang-engineering.com> i2c: remove i2c_new_dummy() API All in-kernel users have been converted to {devm_}i2c_new_dummy_device(). Remove the old API. Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se> Tested-by: Luca Ceresoli <luca@lucaceresoli.net> Reviewed-by: Luca Ceresoli <luca@lucaceresoli.net> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> diff c942fddf Mon May 27 00:55:06 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 157 Based on 3 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version [author] [kishon] [vijay] [abraham] [i] [kishon]@[ti] [com] this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version [author] [graeme] [gregory] [gg]@[slimlogic] [co] [uk] [author] [kishon] [vijay] [abraham] [i] [kishon]@[ti] [com] [based] [on] [twl6030]_[usb] [c] [author] [hema] [hk] [hemahk]@[ti] [com] this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 1105 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Richard Fontana <rfontana@redhat.com> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070033.202006027@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff c942fddf Mon May 27 00:55:06 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 157 Based on 3 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version [author] [kishon] [vijay] [abraham] [i] [kishon]@[ti] [com] this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version [author] [graeme] [gregory] [gg]@[slimlogic] [co] [uk] [author] [kishon] [vijay] [abraham] [i] [kishon]@[ti] [com] [based] [on] [twl6030]_[usb] [c] [author] [hema] [hk] [hemahk]@[ti] [com] this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 1105 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Richard Fontana <rfontana@redhat.com> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070033.202006027@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff c942fddf Mon May 27 00:55:06 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 157 Based on 3 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version [author] [kishon] [vijay] [abraham] [i] [kishon]@[ti] [com] this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version [author] [graeme] [gregory] [gg]@[slimlogic] [co] [uk] [author] [kishon] [vijay] [abraham] [i] [kishon]@[ti] [com] [based] [on] [twl6030]_[usb] [c] [author] [hema] [hk] [hemahk]@[ti] [com] this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 1105 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Richard Fontana <rfontana@redhat.com> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070033.202006027@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff bae1d3a0 Wed Apr 03 06:40:08 MDT 2019 Wolfram Sang <wsa+renesas@sang-engineering.com> i2c: core: remove use of in_atomic() Commit cea443a81c9c ("i2c: Support i2c_transfer in atomic contexts") added in_atomic() to the I2C core. However, the use of in_atomic() outside of core kernel code is discouraged and was already[1] when this code was added in early 2008. The above commit was a preparation for commit b7a3670131c7 ("i2c-pxa: Add polling transfer"). Its commit message says explicitly it was added "for cases where I2C transactions have to occur at times interrup[t]s are disabled". So, the intention was 'disabled interrupts'. This matches the use cases for atomic I2C transfers I have seen so far: very late communication (mostly to a PMIC) to powerdown or reboot the system. For those cases, interrupts are disabled then. It doesn't seem that in_atomic() adds value. After a discussion with Peter Zijlstra[2], we came up with a better set of conditionals to match the use case. The I2C core will soon gain an extra callback into bus drivers especially for atomic transfers to make them more generic. The code deciding which transfer to use (atomic/non-atomic) should mimic the behaviour which locking to use (trylock/lock). This is why we add a helper for it. [1] https://lwn.net/Articles/274695/ [2] http://patchwork.ozlabs.org/patch/1067437/ Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by Andy Shevchenko <andriy.shevchenko@linux.intel.com> Tested-by: Stefan Lengfeld <contact@stefanchrist.eu> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> diff bae1d3a0 Wed Apr 03 06:40:08 MDT 2019 Wolfram Sang <wsa+renesas@sang-engineering.com> i2c: core: remove use of in_atomic() Commit cea443a81c9c ("i2c: Support i2c_transfer in atomic contexts") added in_atomic() to the I2C core. However, the use of in_atomic() outside of core kernel code is discouraged and was already[1] when this code was added in early 2008. The above commit was a preparation for commit b7a3670131c7 ("i2c-pxa: Add polling transfer"). Its commit message says explicitly it was added "for cases where I2C transactions have to occur at times interrup[t]s are disabled". So, the intention was 'disabled interrupts'. This matches the use cases for atomic I2C transfers I have seen so far: very late communication (mostly to a PMIC) to powerdown or reboot the system. For those cases, interrupts are disabled then. It doesn't seem that in_atomic() adds value. After a discussion with Peter Zijlstra[2], we came up with a better set of conditionals to match the use case. The I2C core will soon gain an extra callback into bus drivers especially for atomic transfers to make them more generic. The code deciding which transfer to use (atomic/non-atomic) should mimic the behaviour which locking to use (trylock/lock). This is why we add a helper for it. [1] https://lwn.net/Articles/274695/ [2] http://patchwork.ozlabs.org/patch/1067437/ Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by Andy Shevchenko <andriy.shevchenko@linux.intel.com> Tested-by: Stefan Lengfeld <contact@stefanchrist.eu> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> diff 7b94ea50 Fri Jul 20 02:39:14 MDT 2018 Peter Rosin <peda@axentia.se> i2c/mux, locking/core: Annotate the nested rt_mutex usage If an i2c topology has instances of nested muxes, then a lockdep splat is produced when when i2c_parent_lock_bus() is called. Here is an example: ============================================ WARNING: possible recursive locking detected -------------------------------------------- insmod/68159 is trying to acquire lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] but task is already holding lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(i2c_register_adapter#2); lock(i2c_register_adapter#2); *** DEADLOCK *** May be due to missing lock nesting notation 1 lock held by insmod/68159: #0: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] stack backtrace: CPU: 13 PID: 68159 Comm: insmod Tainted: G O Call Trace: dump_stack+0x67/0x98 __lock_acquire+0x162e/0x1780 lock_acquire+0xba/0x200 rt_mutex_lock+0x44/0x60 i2c_parent_lock_bus+0x32/0x50 [i2c_mux] i2c_parent_lock_bus+0x3e/0x50 [i2c_mux] i2c_smbus_xfer+0xf0/0x700 i2c_smbus_read_byte+0x42/0x70 my2c_init+0xa2/0x1000 [my2c] do_one_initcall+0x51/0x192 do_init_module+0x62/0x216 load_module+0x20f9/0x2b50 SYSC_init_module+0x19a/0x1c0 SyS_init_module+0xe/0x10 do_syscall_64+0x6c/0x1a0 entry_SYSCALL_64_after_hwframe+0x42/0xb7 Reported-by: John Sperbeck <jsperbeck@google.com> Tested-by: John Sperbeck <jsperbeck@google.com> Signed-off-by: Peter Rosin <peda@axentia.se> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Deepa Dinamani <deepadinamani@google.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Chang <dpf@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Cc: Wolfram Sang <wsa@the-dreams.de> Link: http://lkml.kernel.org/r/20180720083914.1950-3-peda@axentia.se Signed-off-by: Ingo Molnar <mingo@kernel.org> diff 7b94ea50 Fri Jul 20 02:39:14 MDT 2018 Peter Rosin <peda@axentia.se> i2c/mux, locking/core: Annotate the nested rt_mutex usage If an i2c topology has instances of nested muxes, then a lockdep splat is produced when when i2c_parent_lock_bus() is called. Here is an example: ============================================ WARNING: possible recursive locking detected -------------------------------------------- insmod/68159 is trying to acquire lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] but task is already holding lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(i2c_register_adapter#2); lock(i2c_register_adapter#2); *** DEADLOCK *** May be due to missing lock nesting notation 1 lock held by insmod/68159: #0: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] stack backtrace: CPU: 13 PID: 68159 Comm: insmod Tainted: G O Call Trace: dump_stack+0x67/0x98 __lock_acquire+0x162e/0x1780 lock_acquire+0xba/0x200 rt_mutex_lock+0x44/0x60 i2c_parent_lock_bus+0x32/0x50 [i2c_mux] i2c_parent_lock_bus+0x3e/0x50 [i2c_mux] i2c_smbus_xfer+0xf0/0x700 i2c_smbus_read_byte+0x42/0x70 my2c_init+0xa2/0x1000 [my2c] do_one_initcall+0x51/0x192 do_init_module+0x62/0x216 load_module+0x20f9/0x2b50 SYSC_init_module+0x19a/0x1c0 SyS_init_module+0xe/0x10 do_syscall_64+0x6c/0x1a0 entry_SYSCALL_64_after_hwframe+0x42/0xb7 Reported-by: John Sperbeck <jsperbeck@google.com> Tested-by: John Sperbeck <jsperbeck@google.com> Signed-off-by: Peter Rosin <peda@axentia.se> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Deepa Dinamani <deepadinamani@google.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Chang <dpf@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Cc: Wolfram Sang <wsa@the-dreams.de> Link: http://lkml.kernel.org/r/20180720083914.1950-3-peda@axentia.se Signed-off-by: Ingo Molnar <mingo@kernel.org> diff 7b94ea50 Fri Jul 20 02:39:14 MDT 2018 Peter Rosin <peda@axentia.se> i2c/mux, locking/core: Annotate the nested rt_mutex usage If an i2c topology has instances of nested muxes, then a lockdep splat is produced when when i2c_parent_lock_bus() is called. Here is an example: ============================================ WARNING: possible recursive locking detected -------------------------------------------- insmod/68159 is trying to acquire lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] but task is already holding lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(i2c_register_adapter#2); lock(i2c_register_adapter#2); *** DEADLOCK *** May be due to missing lock nesting notation 1 lock held by insmod/68159: #0: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] stack backtrace: CPU: 13 PID: 68159 Comm: insmod Tainted: G O Call Trace: dump_stack+0x67/0x98 __lock_acquire+0x162e/0x1780 lock_acquire+0xba/0x200 rt_mutex_lock+0x44/0x60 i2c_parent_lock_bus+0x32/0x50 [i2c_mux] i2c_parent_lock_bus+0x3e/0x50 [i2c_mux] i2c_smbus_xfer+0xf0/0x700 i2c_smbus_read_byte+0x42/0x70 my2c_init+0xa2/0x1000 [my2c] do_one_initcall+0x51/0x192 do_init_module+0x62/0x216 load_module+0x20f9/0x2b50 SYSC_init_module+0x19a/0x1c0 SyS_init_module+0xe/0x10 do_syscall_64+0x6c/0x1a0 entry_SYSCALL_64_after_hwframe+0x42/0xb7 Reported-by: John Sperbeck <jsperbeck@google.com> Tested-by: John Sperbeck <jsperbeck@google.com> Signed-off-by: Peter Rosin <peda@axentia.se> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Deepa Dinamani <deepadinamani@google.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Chang <dpf@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Cc: Wolfram Sang <wsa@the-dreams.de> Link: http://lkml.kernel.org/r/20180720083914.1950-3-peda@axentia.se Signed-off-by: Ingo Molnar <mingo@kernel.org> diff 7b94ea50 Fri Jul 20 02:39:14 MDT 2018 Peter Rosin <peda@axentia.se> i2c/mux, locking/core: Annotate the nested rt_mutex usage If an i2c topology has instances of nested muxes, then a lockdep splat is produced when when i2c_parent_lock_bus() is called. Here is an example: ============================================ WARNING: possible recursive locking detected -------------------------------------------- insmod/68159 is trying to acquire lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] but task is already holding lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(i2c_register_adapter#2); lock(i2c_register_adapter#2); *** DEADLOCK *** May be due to missing lock nesting notation 1 lock held by insmod/68159: #0: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] stack backtrace: CPU: 13 PID: 68159 Comm: insmod Tainted: G O Call Trace: dump_stack+0x67/0x98 __lock_acquire+0x162e/0x1780 lock_acquire+0xba/0x200 rt_mutex_lock+0x44/0x60 i2c_parent_lock_bus+0x32/0x50 [i2c_mux] i2c_parent_lock_bus+0x3e/0x50 [i2c_mux] i2c_smbus_xfer+0xf0/0x700 i2c_smbus_read_byte+0x42/0x70 my2c_init+0xa2/0x1000 [my2c] do_one_initcall+0x51/0x192 do_init_module+0x62/0x216 load_module+0x20f9/0x2b50 SYSC_init_module+0x19a/0x1c0 SyS_init_module+0xe/0x10 do_syscall_64+0x6c/0x1a0 entry_SYSCALL_64_after_hwframe+0x42/0xb7 Reported-by: John Sperbeck <jsperbeck@google.com> Tested-by: John Sperbeck <jsperbeck@google.com> Signed-off-by: Peter Rosin <peda@axentia.se> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Deepa Dinamani <deepadinamani@google.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Chang <dpf@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Cc: Wolfram Sang <wsa@the-dreams.de> Link: http://lkml.kernel.org/r/20180720083914.1950-3-peda@axentia.se Signed-off-by: Ingo Molnar <mingo@kernel.org> diff 7b94ea50 Fri Jul 20 02:39:14 MDT 2018 Peter Rosin <peda@axentia.se> i2c/mux, locking/core: Annotate the nested rt_mutex usage If an i2c topology has instances of nested muxes, then a lockdep splat is produced when when i2c_parent_lock_bus() is called. Here is an example: ============================================ WARNING: possible recursive locking detected -------------------------------------------- insmod/68159 is trying to acquire lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] but task is already holding lock: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(i2c_register_adapter#2); lock(i2c_register_adapter#2); *** DEADLOCK *** May be due to missing lock nesting notation 1 lock held by insmod/68159: #0: (i2c_register_adapter#2){+.+.}, at: i2c_parent_lock_bus+0x32/0x50 [i2c_mux] stack backtrace: CPU: 13 PID: 68159 Comm: insmod Tainted: G O Call Trace: dump_stack+0x67/0x98 __lock_acquire+0x162e/0x1780 lock_acquire+0xba/0x200 rt_mutex_lock+0x44/0x60 i2c_parent_lock_bus+0x32/0x50 [i2c_mux] i2c_parent_lock_bus+0x3e/0x50 [i2c_mux] i2c_smbus_xfer+0xf0/0x700 i2c_smbus_read_byte+0x42/0x70 my2c_init+0xa2/0x1000 [my2c] do_one_initcall+0x51/0x192 do_init_module+0x62/0x216 load_module+0x20f9/0x2b50 SYSC_init_module+0x19a/0x1c0 SyS_init_module+0xe/0x10 do_syscall_64+0x6c/0x1a0 entry_SYSCALL_64_after_hwframe+0x42/0xb7 Reported-by: John Sperbeck <jsperbeck@google.com> Tested-by: John Sperbeck <jsperbeck@google.com> Signed-off-by: Peter Rosin <peda@axentia.se> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Davidlohr Bueso <dave@stgolabs.net> Cc: Deepa Dinamani <deepadinamani@google.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Chang <dpf@google.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Philippe Ombredanne <pombredanne@nexb.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Cc: Wolfram Sang <wsa@the-dreams.de> Link: http://lkml.kernel.org/r/20180720083914.1950-3-peda@axentia.se Signed-off-by: Ingo Molnar <mingo@kernel.org> |
/linux-master/include/linux/ | ||
H A D | i2c.h | diff 2f6a470d Mon Nov 15 08:49:46 MST 2021 Jakub Kicinski <kuba@kernel.org> Revert "Merge branch 'mctp-i2c-driver'" This reverts commit 71812af7234f30362b43ccff33f93890ae4c0655, reversing changes made to cc0be1ad686fb29a4d127948486f40b17fb34b50. Wolfram Sang says: Please revert. Besides the driver in net, it modifies the I2C core code. This has not been acked by the I2C maintainer (in this case me). So, please don't pull this in via the net tree. The question raised here (extending SMBus calls to 255 byte) is complicated because we need ABI backwards compatibility. Link: https://lore.kernel.org/all/YZJ9H4eM%2FM7OXVN0@shikoro/ Signed-off-by: Jakub Kicinski <kuba@kernel.org> diff 2f6a470d Mon Nov 15 08:49:46 MST 2021 Jakub Kicinski <kuba@kernel.org> Revert "Merge branch 'mctp-i2c-driver'" This reverts commit 71812af7234f30362b43ccff33f93890ae4c0655, reversing changes made to cc0be1ad686fb29a4d127948486f40b17fb34b50. Wolfram Sang says: Please revert. Besides the driver in net, it modifies the I2C core code. This has not been acked by the I2C maintainer (in this case me). So, please don't pull this in via the net tree. The question raised here (extending SMBus calls to 255 byte) is complicated because we need ABI backwards compatibility. Link: https://lore.kernel.org/all/YZJ9H4eM%2FM7OXVN0@shikoro/ Signed-off-by: Jakub Kicinski <kuba@kernel.org> diff 20a1b3ac Tue Aug 03 10:00:41 MDT 2021 Hans de Goede <hdegoede@redhat.com> i2c: acpi: Add an i2c_acpi_client_count() helper function We have 3 files now which have the need to count the number of I2cSerialBus resources in an ACPI-device's resource-list. Currently all implement their own helper function for this, add a generic helper function to replace the 3 implementations. Signed-off-by: Hans de Goede <hdegoede@redhat.com> Link: https://lore.kernel.org/r/20210803160044.158802-2-hdegoede@redhat.com Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com> Acked-by: Wolfram Sang <wsa@kernel.org> diff 8be23aec Sat Jul 25 08:07:36 MDT 2020 Wolfram Sang <wsa+renesas@sang-engineering.com> i2c: also convert placeholder function to return errno All i2c_new_device-alike functions return ERR_PTR these days, but this fallback function was missed. Fixes: 2dea645ffc21 ("i2c: acpi: Return error pointers from i2c_acpi_new_device()") Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com> [wsa: changed from 'ENOSYS' to 'ENODEV'] Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2f5a55c5 Sat May 02 06:18:35 MDT 2020 Wolfram Sang <wsa@kernel.org> i2c: use my kernel.org address from now on The old email is still active, but for easier handling, I am going to use my kernel.org address from now on. Also, add a mailmap for the now defunct Pengutronix address. Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2f5a55c5 Sat May 02 06:18:35 MDT 2020 Wolfram Sang <wsa@kernel.org> i2c: use my kernel.org address from now on The old email is still active, but for easier handling, I am going to use my kernel.org address from now on. Also, add a mailmap for the now defunct Pengutronix address. Signed-off-by: Wolfram Sang <wsa@kernel.org> diff 2c2f00ab Thu Dec 05 17:23:22 MST 2019 Wolfram Sang <wsa+renesas@sang-engineering.com> i2c: remove i2c_new_dummy() API All in-kernel users have been converted to {devm_}i2c_new_dummy_device(). Remove the old API. Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Reviewed-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se> Tested-by: Luca Ceresoli <luca@lucaceresoli.net> Reviewed-by: Luca Ceresoli <luca@lucaceresoli.net> Signed-off-by: Wolfram Sang <wsa@the-dreams.de> diff 2caea56f Mon Jun 03 02:25:34 MDT 2019 Wolfram Sang <wsa@kernel.org> i2c: headers: update docs about I2C_CLIENT_* Update kerneldoc for i2c client flags because they increased over time. Also, move them to a position where they can be more easily found. Signed-off-by: Wolfram Sang <wsa@the-dreams.de> diff 16216333 Sun May 19 07:51:31 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 1 Based on 2 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin street fifth floor boston ma 02110 1301 usa this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option [no]_[pad]_[ctrl] any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin street fifth floor boston ma 02110 1301 usa extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 176 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com> Reviewed-by: Steve Winslow <swinslow@gmail.com> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190519154040.652910950@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff 16216333 Sun May 19 07:51:31 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 1 Based on 2 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin street fifth floor boston ma 02110 1301 usa this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option [no]_[pad]_[ctrl] any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin street fifth floor boston ma 02110 1301 usa extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 176 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com> Reviewed-by: Steve Winslow <swinslow@gmail.com> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190519154040.652910950@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff 16216333 Sun May 19 07:51:31 MDT 2019 Thomas Gleixner <tglx@linutronix.de> treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 1 Based on 2 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin street fifth floor boston ma 02110 1301 usa this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option [no]_[pad]_[ctrl] any later version this program is distributed in the hope that it will be useful but without any warranty without even the implied warranty of merchantability or fitness for a particular purpose see the gnu general public license for more details you should have received a copy of the gnu general public license along with this program if not write to the free software foundation inc 51 franklin street fifth floor boston ma 02110 1301 usa extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 176 file(s). Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Jilayne Lovejoy <opensource@jilayne.com> Reviewed-by: Steve Winslow <swinslow@gmail.com> Reviewed-by: Allison Randal <allison@lohutok.net> Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org> Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190519154040.652910950@linutronix.de Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> |
/linux-master/ | ||
H A D | .mailmap | diff 71cfc131 Fri Feb 02 11:23:39 MST 2024 Nadav Amit <nadav.amit@gmail.com> vmw_balloon: change maintainership Jerrin will be the new maintainer of the VMware balloon driver following Broadcom's acquisition and Nadav's departure. Update accordingly: 1. Update the maintainer name and email. 2. Update the reviewer list to Broadcom's, which acquired VMware. 3. Add .mailmap entries for Nadav. Cc: Jerrin Shaji George <jerrin.shaji-george@broadcom.com> Signed-off-by: Nadav Amit <nadav.amit@gmail.com> Acked-by: Jerrin Shaji George <jerrin.shaji-george@broadcom.com> Link: https://lore.kernel.org/r/20240202182339.1725466-1-nadav.amit@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff 2d880bfa Mon Dec 04 02:37:15 MST 2023 Jiri Kosina <jikos@kernel.org> mailmap: add address mapping for Jiri Kosina Since I switched the MAINTAINERS entry to @kernel.org some time ago, let's canonicalize my addressess to it. Signed-off-by: Jiri Kosina <jikos@kernel.org> Signed-off-by: Jiri Kosina <jkosina@suse.com> diff 4aa8f278 Mon Oct 30 19:40:00 MDT 2023 Bagas Sanjaya <bagasdotme@gmail.com> .mailmap: add address mapping for Tomeu Vizoso He's no longer working in Collabora (and his email address there bounces). Map it to his personal address. Link: https://lkml.kernel.org/r/20231031014009.22765-2-bagasdotme@gmail.com Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com> Acked-by: Tomeu Vizoso <tomeu@tomeuvizoso.net> Cc: Bjorn Andersson <quic_bjorande@quicinc.com> Cc: Heiko Stuebner <heiko@sntech.de> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Jens Axboe <axboe@kernel.dk> Cc: Konrad Dybcio <konrad.dybcio@linaro.org> Cc: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 90723a82 Mon Oct 30 08:24:55 MDT 2023 Bagas Sanjaya <bagasdotme@gmail.com> .mailmap: map Benjamin Poirier's address Map out to his gmail address as he had left SUSE some time ago. Link: https://lkml.kernel.org/r/20231030142454.22127-2-bagasdotme@gmail.com Signed-off-by: Bagas Sanjaya <bagasdotme@gmail.com> Acked-by: Benjamin Poirier <benjamin.poirier@gmail.com> Cc: Bjorn Andersson <quic_bjorande@quicinc.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Heiko Stuebner <heiko@sntech.de> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Konrad Dybcio <konrad.dybcio@linaro.org> Cc: Oleksij Rempel <o.rempel@pengutronix.de> Cc: Stephen Hemminger <stephen@networkplumber.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff 54ccc875 Sat Jul 08 13:53:09 MDT 2023 Jonas Gorski <jonas.gorski@gmail.com> mailmap: add entry for Jonas Gorski The openwrt.org email address is long defunct, but still pop ups from time to time when asking get_maintainer.pl. So add an entry to my currently used address. Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com> Link: https://lore.kernel.org/r/20230708195309.72767-2-jonas.gorski@gmail.com Signed-off-by: Mark Brown <broonie@kernel.org> diff ddcd91f4 Tue Jul 04 10:39:18 MDT 2023 Heiko Stuebner <heiko@sntech.de> mailmap: update manpage link Patch series "Update .mailmap for my work address and fix manpage". While updating mailmap for the going-away address, I also found that on current systems the manpage linked from the header comment changed. And in fact it looks like the git mailmap feature got its own manpage. This patch (of 2): On recent systems the git-shortlog manpage only tells people to See gitmailmap(5) So instead of sending people on a scavenger hunt, put that info into the header directly. Though keep the old reference around for older systems. Link: https://lkml.kernel.org/r/20230704163919.1136784-1-heiko@sntech.de Link: https://lkml.kernel.org/r/20230704163919.1136784-2-heiko@sntech.de Signed-off-by: Heiko Stuebner <heiko.stuebner@vrull.eu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff ddcd91f4 Tue Jul 04 10:39:18 MDT 2023 Heiko Stuebner <heiko@sntech.de> mailmap: update manpage link Patch series "Update .mailmap for my work address and fix manpage". While updating mailmap for the going-away address, I also found that on current systems the manpage linked from the header comment changed. And in fact it looks like the git mailmap feature got its own manpage. This patch (of 2): On recent systems the git-shortlog manpage only tells people to See gitmailmap(5) So instead of sending people on a scavenger hunt, put that info into the header directly. Though keep the old reference around for older systems. Link: https://lkml.kernel.org/r/20230704163919.1136784-1-heiko@sntech.de Link: https://lkml.kernel.org/r/20230704163919.1136784-2-heiko@sntech.de Signed-off-by: Heiko Stuebner <heiko.stuebner@vrull.eu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff ecf1d926 Tue Feb 28 08:33:34 MST 2023 Konrad Dybcio <konrad.dybcio@linaro.org> mailmap: map Dikshita Agarwal's old address to his current one Dikshita's old email is still picked up by the likes of get_maintainer.pl and keeps bouncing. Map it to his current one. Link: https://lkml.kernel.org/r/20230228153335.907164-2-konrad.dybcio@linaro.org Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org> Cc: Dikshita Agarwal <dikshita@qti.qualcomm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> diff f1ab2e0d Tue Nov 16 06:58:03 MST 2021 Christian Borntraeger <borntraeger@linux.ibm.com> MAINTAINERS: update email address of Christian Borntraeger My borntraeger@de.ibm.com email is just a forwarder to the linux.ibm.com address. Let us remove the extra hop to avoid a potential source of errors. While at it, add the relevant email addresses to mailmap. Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com> Link: https://lore.kernel.org/r/20211116135803.119489-2-borntraeger@linux.ibm.com Signed-off-by: Heiko Carstens <hca@linux.ibm.com> diff 72a461ad Thu Jun 24 19:40:13 MDT 2021 Marek Behún <kabel@kernel.org> mailmap: add Marek's other e-mail address and identity without diacritics Some of my commits were sent with identities Marek Behun <marek.behun@nic.cz> Marek Behún <marek.behun@nic.cz> while the correct one is Marek Behún <kabel@kernel.org> Put this into mailmap so that git shortlog prints all my commits under one identity. Link: https://lkml.kernel.org/r/20210616113624.19351-2-kabel@kernel.org Signed-off-by: Marek Behún <kabel@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> |
H A D | MAINTAINERS | diff 71cfc131 Fri Feb 02 11:23:39 MST 2024 Nadav Amit <nadav.amit@gmail.com> vmw_balloon: change maintainership Jerrin will be the new maintainer of the VMware balloon driver following Broadcom's acquisition and Nadav's departure. Update accordingly: 1. Update the maintainer name and email. 2. Update the reviewer list to Broadcom's, which acquired VMware. 3. Add .mailmap entries for Nadav. Cc: Jerrin Shaji George <jerrin.shaji-george@broadcom.com> Signed-off-by: Nadav Amit <nadav.amit@gmail.com> Acked-by: Jerrin Shaji George <jerrin.shaji-george@broadcom.com> Link: https://lore.kernel.org/r/20240202182339.1725466-1-nadav.amit@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> diff 5659785a Tue Jan 23 01:10:59 MST 2024 Kim Seer Paller <kimseer.paller@analog.com> iio: frequency: admfm2000: New driver Dual microwave down converter module with input RF and LO frequency ranges from 0.5 to 32 GHz and an output IF frequency range from 0.1 to 8 GHz. It consists of a LNA, mixer, IF filter, DSA, and IF amplifier for each down conversion path. Signed-off-by: Kim Seer Paller <kimseer.paller@analog.com> Link: https://lore.kernel.org/r/20240123081059.5746-2-kimseer.paller@analog.com Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com> diff d6cfd177 Wed Jan 31 07:49:33 MST 2024 Andrea Parri <parri.andrea@gmail.com> membarrier: riscv: Add full memory barrier in switch_mm() The membarrier system call requires a full memory barrier after storing to rq->curr, before going back to user-space. The barrier is only needed when switching between processes: the barrier is implied by mmdrop() when switching from kernel to userspace, and it's not needed when switching from userspace to kernel. Rely on the feature/mechanism ARCH_HAS_MEMBARRIER_CALLBACKS and on the primitive membarrier_arch_switch_mm(), already adopted by the PowerPC architecture, to insert the required barrier. Fixes: fab957c11efe2f ("RISC-V: Atomic and Locking Code") Signed-off-by: Andrea Parri <parri.andrea@gmail.com> Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Link: https://lore.kernel.org/r/20240131144936.29190-2-parri.andrea@gmail.com Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com> diff c5839535 Thu Feb 22 10:43:42 MST 2024 Rob Herring <robh@kernel.org> dt-bindings: i2c: Remove obsolete i2c.txt Everything in i2c.txt is covered by schemas/i2c/i2c-controller.yaml in dtschema project, so remove i2c.txt and update links to it in the tree. Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com> Acked-by: Mark Brown <broonie@kernel.org> Link: https://lore.kernel.org/r/20240222174343.3482354-2-robh@kernel.org Signed-off-by: Rob Herring <robh@kernel.org> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> diff 701ab935 Tue Feb 20 04:50:55 MST 2024 Ankit Agrawal <ankita@nvidia.com> vfio/nvgrace-gpu: Add vfio pci variant module for grace hopper NVIDIA's upcoming Grace Hopper Superchip provides a PCI-like device for the on-chip GPU that is the logical OS representation of the internal proprietary chip-to-chip cache coherent interconnect. The device is peculiar compared to a real PCI device in that whilst there is a real 64b PCI BAR1 (comprising region 2 & region 3) on the device, it is not used to access device memory once the faster chip-to-chip interconnect is initialized (occurs at the time of host system boot). The device memory is accessed instead using the chip-to-chip interconnect that is exposed as a contiguous physically addressable region on the host. This device memory aperture can be obtained from host ACPI table using device_property_read_u64(), according to the FW specification. Since the device memory is cache coherent with the CPU, it can be mmap into the user VMA with a cacheable mapping using remap_pfn_range() and used like a regular RAM. The device memory is not added to the host kernel, but mapped directly as this reduces memory wastage due to struct pages. There is also a requirement of a minimum reserved 1G uncached region (termed as resmem) to support the Multi-Instance GPU (MIG) feature [1]. This is to work around a HW defect. Based on [2], the requisite properties (uncached, unaligned access) can be achieved through a VM mapping (S1) of NORMAL_NC and host (S2) mapping with MemAttr[2:0]=0b101. To provide a different non-cached property to the reserved 1G region, it needs to be carved out from the device memory and mapped as a separate region in Qemu VMA with pgprot_writecombine(). pgprot_writecombine() sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Provide a VFIO PCI variant driver that adapts the unique device memory representation into a more standard PCI representation facing userspace. The variant driver exposes these two regions - the non-cached reserved (resmem) and the cached rest of the device memory (termed as usemem) as separate VFIO 64b BAR regions. This is divergent from the baremetal approach, where the device memory is exposed as a device memory region. The decision for a different approach was taken in view of the fact that it would necessiate additional code in Qemu to discover and insert those regions in the VM IPA, along with the additional VM ACPI DSDT changes to communicate the device memory region IPA to the VM workloads. Moreover, this behavior would have to be added to a variety of emulators (beyond top of tree Qemu) out there desiring grace hopper support. Since the device implements 64-bit BAR0, the VFIO PCI variant driver maps the uncached carved out region to the next available PCI BAR (i.e. comprising of region 2 and 3). The cached device memory aperture is assigned BAR region 4 and 5. Qemu will then naturally generate a PCI device in the VM with the uncached aperture reported as BAR2 region, the cacheable as BAR4. The variant driver provides emulation for these fake BARs' PCI config space offset registers. The hardware ensures that the system does not crash when the memory is accessed with the memory enable turned off. It synthesis ~0 reads and dropped writes on such access. So there is no need to support the disablement/enablement of BAR through PCI_COMMAND config space register. The memory layout on the host looks like the following: devmem (memlength) |--------------------------------------------------| |-------------cached------------------------|--NC--| | | usemem.memphys resmem.memphys PCI BARs need to be aligned to the power-of-2, but the actual memory on the device may not. A read or write access to the physical address from the last device PFN up to the next power-of-2 aligned physical address results in reading ~0 and dropped writes. Note that the GPU device driver [6] is capable of knowing the exact device memory size through separate means. The device memory size is primarily kept in the system ACPI tables for use by the VFIO PCI variant module. Note that the usemem memory is added by the VM Nvidia device driver [5] to the VM kernel as memblocks. Hence make the usable memory size memblock (MEMBLK_SIZE) aligned. This is a hardwired ABI value between the GPU FW and VFIO driver. The VM device driver make use of the same value for its calculation to determine USEMEM size. Currently there is no provision in KVM for a S2 mapping with MemAttr[2:0]=0b101, but there is an ongoing effort to provide the same [3]. As previously mentioned, resmem is mapped pgprot_writecombine(), that sets the Qemu VMA page properties (pgprot) as NORMAL_NC. Using the proposed changes in [3] and [4], KVM marks the region with MemAttr[2:0]=0b101 in S2. If the device memory properties are not present, the driver registers the vfio-pci-core function pointers. Since there are no ACPI memory properties generated for the VM, the variant driver inside the VM will only use the vfio-pci-core ops and hence try to map the BARs as non cached. This is not a problem as the CPUs have FWB enabled which blocks the VM mapping's ability to override the cacheability set by the host mapping. This goes along with a qemu series [6] to provides the necessary implementation of the Grace Hopper Superchip firmware specification so that the guest operating system can see the correct ACPI modeling for the coherent GPU device. Verified with the CUDA workload in the VM. [1] https://www.nvidia.com/en-in/technologies/multi-instance-gpu/ [2] section D8.5.5 of https://developer.arm.com/documentation/ddi0487/latest/ [3] https://lore.kernel.org/all/20240211174705.31992-1-ankita@nvidia.com/ [4] https://lore.kernel.org/all/20230907181459.18145-2-ankita@nvidia.com/ [5] https://github.com/NVIDIA/open-gpu-kernel-modules [6] https://lore.kernel.org/all/20231203060245.31593-1-ankita@nvidia.com/ Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Yishai Hadas <yishaih@nvidia.com> Reviewed-by: Zhi Wang <zhi.wang.linux@gmail.com> Signed-off-by: Aniket Agashe <aniketa@nvidia.com> Signed-off-by: Ankit Agrawal <ankita@nvidia.com> Link: https://lore.kernel.org/r/20240220115055.23546-4-ankita@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> |
Completed in 2609 milliseconds