Clean things up, refrain from making changes that may cause unnecessary complications in the future

Revert "Remove redundant cache code, rename files, tie everything in"

This reverts commit f96a540c97eb63b9360fb0c254869cd745c8ac99.

Remove redundant cache code, rename files, tie everything in

Eliminate most compiler warnings in pmap, fix nasty bug with cpm_allocate()

Implement more stubs, implement a temporary workaround for the lack of corecrypto.kext so SHA1* and MD5* function calls don't have to be gated.

This is @superMTW's fault.

Preliminary fixes for KASLR support.

ASID support.

Do not mess with page entries if they're outside of RAM.

ARM11 support, not tested on hardware. Compiles using Xcode 4.4.1.

Very experimental -O2 kernel config, doesn't seem to work right with Linux clang, works fine with clang from Xcode 4.6.2.

Every page is now backed by a lock on it. Only one operation may be done on a pv_h at one time.

Print out 'shadow' page table bits and fix a typo.

Lock/unlock pv_hs as required (really just enable/disable preemption for right now.)

Implement pmap cacheability and cacheability masks. Sort of incomplete at the moment though.

High order addresses are just invalid, not fatal. Fixes syslogd.

KDP will now work properly again.

This doesn't necessarily always work so well...

One day, I promise to stop modifying this file.

indent never likes this statement.

Reformatting only.

Properly free expanded pages when needed. This resolves the biggest memory leak in the system.

More pmap improvements, add bookkeepping for pt_zone_info, add sysctl parameters for machdep.pmap.

Simplify pmap_enter_options and friends, use similar code flow as per i386 for all of the internal pmap bookkeeping.

Misc. pmap improvements: enable ledger credits/debits, set additional machine dependent bits in the pv_rooted_entry table.

Implement faster pmap_extract, this one uses the ARM PAR and c7 PA to VA translation register.

General cleanup of the codebase, remove redundant files. You will need to do a full rebuild after this commit.

Reduce memory burning.

Fix a typo.

For new pages, add cacheability bits.

Fix... launchd and all CoreFoundation/IOKit programs.

PTE hits for the commonpage and later are invalid for the user pmap.

Reduce calls to splx/splhigh and replace them with spinlocks.

bash works now.

ASID support and misc. pmap fixes... (again)

Add support for 'dataconstro', force all CONST data to remain as RO and NoExecute.

Make the kernel text R-X by default.

Switch pmap over from the old linear tree mapping code to the pv_hash code using in xnu-1228's i386 pmap. This should improve performance. Additionally, clear only specific regions of the TLB as required instead of the full TLB.

pmap cleanup.

Reformatting.

pmap_expand is now fixed, but I have some hardware issues to debug with cpm_allocate/kmem_alloc_contig.

Implement pmap_nest/pmap_unnest. These functions are used for shared segments on ARM. pmap_expand has to be fixed to also work properly.

Userspace and pmap fixes. Warning: Userspace is still buggy, kernel dies after about 4 forks.

Attempt to fix more of pmap.

Implement pmap_protect.

Reformat everything to fit in an 80 column setup, makes certain functions look far less scary.

Remove the bringup hack, this fixes launchd zombies, but now in internal launchd logs I see vm_allocate(mach_task_self(), (vm_address_t *)&req, mz, flags) == KERN_SUCCESS) failing.

Add full support for NX, and VM protections.

Fix launchd[2] fork issue, now pmap needs to be fixed again. Additionally, fix ml_cpu_up and ml_cpu_down. Fixes hostinfo, we now have more than 0 processors in the system.

Hack to make ARM kexts work for right now...

Preliminary support for UNIX signals.

pmap fixes, pthreads work, and assertions for thread initialization/implementation.

*** launchd[1] has started up. ***

Some fixes for booting on hardware, work around a memory corruption issue for now.

Fix uninitialized ledger bug, kill off 8930 hack.

Lock the pmap subsystem on changes, SMP support coming in the future, eventually? Also a very *stupid* bug in sleh_undef.

Silence some warnings, some more sanity checks and processor identification.

Reformatting changes.

Get out of tons of growing pains. Load/store exclusive instructions work on OMAP3530 with caching enabled. Fix that bug and tons more.

pmap_nest?

Refactor arm_vm_init and make vm_map use pmap_nest (which is unimplemented.)

ARM64 build fixes, rewritten Debugger() code. Use nosym=1 and debug= to make the panic log like iOS, otherwise only the context fp/lrs will be printed out and they will be symbolicated.

New 'Halt_system' function, used in panic() to reduce processor energy used.

Some proper fixes, todo, get rid of stupid PAGE_SHIFTs everywhere.

Rewritten sleh_abort.

Today's work.

New pmap, sort of broken though. I'll keep working on it.

DEVELOPMENT kernels work, sorta...

Fix one annoying bug that's been pestering me for a while.

Revert "Lots of random fixes."

This reverts commit c47e4a04153af96ef2703c9eb3e60f753ddb7dc4.

Lots of random fixes.

[OSFMK] Make pmap use WRITE_PTE macro.

Allow I-cache to now be used. Massive *massive* speedup.

[OSFMK]
- Enable caching on entered pmap pages. Still need to fix
i-cache issues on very initial mapping. Things break.
- Implement ml_static_mfree with a stub.
[PEXPERT]
- S5L8930X timer support. (I don't even. This thing needs a TRM.)
- Now we get to IOKit init on this platform, but it crashes again
in atomics...

I changed my email address.

[OSFMK]
- Remove pmap_find_phys, was a return 0 anyway. Should probably be redone.
- Implement pmap_extract to return physical address for a virtual one.
- ml_nofault_copy is a lie.

[IOKIT]
- AppleARMPE now publishes the model name.

Initial kernel import. So many things are broken with this thing, who's willing to fix them all?

Clean things up, refrain from making changes that may cause unnecessary complications in the future

Revert "Remove redundant cache code, rename files, tie everything in"

This reverts commit f96a540c97eb63b9360fb0c254869cd745c8ac99.

Remove redundant cache code, rename files, tie everything in

Eliminate most compiler warnings in pmap, fix nasty bug with cpm_allocate()

Implement more stubs, implement a temporary workaround for the lack of corecrypto.kext so SHA1* and MD5* function calls don't have to be gated.

This is @superMTW's fault.

Preliminary fixes for KASLR support.

ASID support.

Do not mess with page entries if they're outside of RAM.

ARM11 support, not tested on hardware. Compiles using Xcode 4.4.1.

Very experimental -O2 kernel config, doesn't seem to work right with Linux clang, works fine with clang from Xcode 4.6.2.

Every page is now backed by a lock on it. Only one operation may be done on a pv_h at one time.

Print out 'shadow' page table bits and fix a typo.

Lock/unlock pv_hs as required (really just enable/disable preemption for right now.)

Implement pmap cacheability and cacheability masks. Sort of incomplete at the moment though.

High order addresses are just invalid, not fatal. Fixes syslogd.

KDP will now work properly again.

This doesn't necessarily always work so well...

One day, I promise to stop modifying this file.

indent never likes this statement.

Reformatting only.

Properly free expanded pages when needed. This resolves the biggest memory leak in the system.

More pmap improvements, add bookkeepping for pt_zone_info, add sysctl parameters for machdep.pmap.

Simplify pmap_enter_options and friends, use similar code flow as per i386 for all of the internal pmap bookkeeping.

Misc. pmap improvements: enable ledger credits/debits, set additional machine dependent bits in the pv_rooted_entry table.

Implement faster pmap_extract, this one uses the ARM PAR and c7 PA to VA translation register.

General cleanup of the codebase, remove redundant files. You will need to do a full rebuild after this commit.

Reduce memory burning.

Fix a typo.

For new pages, add cacheability bits.

Fix... launchd and all CoreFoundation/IOKit programs.

PTE hits for the commonpage and later are invalid for the user pmap.

Reduce calls to splx/splhigh and replace them with spinlocks.

bash works now.

ASID support and misc. pmap fixes... (again)

Add support for 'dataconstro', force all CONST data to remain as RO and NoExecute.

Make the kernel text R-X by default.

Switch pmap over from the old linear tree mapping code to the pv_hash code using in xnu-1228's i386 pmap. This should improve performance. Additionally, clear only specific regions of the TLB as required instead of the full TLB.

pmap cleanup.

Reformatting.

pmap_expand is now fixed, but I have some hardware issues to debug with cpm_allocate/kmem_alloc_contig.

Implement pmap_nest/pmap_unnest. These functions are used for shared segments on ARM. pmap_expand has to be fixed to also work properly.

Userspace and pmap fixes. Warning: Userspace is still buggy, kernel dies after about 4 forks.

Attempt to fix more of pmap.

Implement pmap_protect.

Reformat everything to fit in an 80 column setup, makes certain functions look far less scary.

Remove the bringup hack, this fixes launchd zombies, but now in internal launchd logs I see vm_allocate(mach_task_self(), (vm_address_t *)&req, mz, flags) == KERN_SUCCESS) failing.

Add full support for NX, and VM protections.

Fix launchd[2] fork issue, now pmap needs to be fixed again. Additionally, fix ml_cpu_up and ml_cpu_down. Fixes hostinfo, we now have more than 0 processors in the system.

Hack to make ARM kexts work for right now...

Preliminary support for UNIX signals.

pmap fixes, pthreads work, and assertions for thread initialization/implementation.

*** launchd[1] has started up. ***

Some fixes for booting on hardware, work around a memory corruption issue for now.

Fix uninitialized ledger bug, kill off 8930 hack.

Lock the pmap subsystem on changes, SMP support coming in the future, eventually? Also a very *stupid* bug in sleh_undef.

Silence some warnings, some more sanity checks and processor identification.

Reformatting changes.

Get out of tons of growing pains. Load/store exclusive instructions work on OMAP3530 with caching enabled. Fix that bug and tons more.

pmap_nest?

Refactor arm_vm_init and make vm_map use pmap_nest (which is unimplemented.)

ARM64 build fixes, rewritten Debugger() code. Use nosym=1 and debug= to make the panic log like iOS, otherwise only the context fp/lrs will be printed out and they will be symbolicated.

New 'Halt_system' function, used in panic() to reduce processor energy used.

Some proper fixes, todo, get rid of stupid PAGE_SHIFTs everywhere.

Rewritten sleh_abort.

Today's work.

New pmap, sort of broken though. I'll keep working on it.

DEVELOPMENT kernels work, sorta...

Fix one annoying bug that's been pestering me for a while.

Revert "Lots of random fixes."

This reverts commit c47e4a04153af96ef2703c9eb3e60f753ddb7dc4.

Lots of random fixes.

[OSFMK] Make pmap use WRITE_PTE macro.

Allow I-cache to now be used. Massive *massive* speedup.

[OSFMK]
- Enable caching on entered pmap pages. Still need to fix
i-cache issues on very initial mapping. Things break.
- Implement ml_static_mfree with a stub.
[PEXPERT]
- S5L8930X timer support. (I don't even. This thing needs a TRM.)
- Now we get to IOKit init on this platform, but it crashes again
in atomics...

I changed my email address.

[OSFMK]
- Remove pmap_find_phys, was a return 0 anyway. Should probably be redone.
- Implement pmap_extract to return physical address for a virtual one.
- ml_nofault_copy is a lie.

[IOKIT]
- AppleARMPE now publishes the model name.

Initial kernel import. So many things are broken with this thing, who's willing to fix them all?