Commit graph

388 commits

Author SHA1 Message Date
Thomas Faber 0af3689c2e
[REACTOS] Fix traces with missing arguments.
Courtesy of VS Code Analysis warning C6064:
Missing integer argument to 'DbgPrint' that corresponds to conversion specifier 'N'.
2021-11-21 12:57:35 -05:00
Thomas Faber a7b2703479
[NTOS] Fix broken SAL annotations on MmDereferenceSegmentWithLock. 2021-11-21 12:47:46 -05:00
Hermès Bélusca-Maïto fe9ac14aa3
[NTOS] Move two CODE_SEG("INIT") to a better place. 2021-09-05 21:22:45 +02:00
Victor Perevertkin 6ef6fabfc5
[FREELDR][NTOS][HALPPC][SDK] Remove PowerPC code
Remove PowerPC-related code from the kernel, HAL, SDK and
Freeloader.
2021-08-15 15:35:51 +03:00
Jérôme Gardou 589016ddb9 [NTOS:MM] Implement MmFlushImageSection(MmFlushForWrite) 2021-08-07 09:34:58 +02:00
Jérôme Gardou acf28dbc1e [NTOS:MM] Fix the mess created by the "balancer thread"
When processing:
  Make sure that the process is not terminating.
  Make sure that the process WorkingSet is still valid
  Protect accessing & writing to PTEs by acquiring the working set lock

CORE-17595 CORE-17642
2021-08-05 10:06:19 +02:00
Jérôme Gardou 676dc4b441 Revert "[NTOS:MM] Do proper checks & acquire proper locks when sweeping user pages"
This reverts commit 7c006df7e0.
Note to self: This is what happens when you have too many branch and push the wrong one.
2021-08-04 23:57:39 +02:00
Jérôme Gardou 7c006df7e0 [NTOS:MM] Do proper checks & acquire proper locks when sweeping user pages
CORE-17595
2021-08-04 20:38:37 +02:00
Jérôme Gardou 789d7a7741 [NTOS:MM] When mapping pages for zeroing them, disable cache, enable write through
Idea from Thomas Faber
2021-08-04 17:48:39 +02:00
Jérôme Gardou 98e62b9bca [NTOS:MM] In page zeroing thread, zero more than one page at a time. 2021-08-04 17:48:39 +02:00
Jérôme Gardou e98de5ef1f [NTOS:MM] Use KeZeroPages in zeroing thread 2021-08-04 17:48:39 +02:00
Jérôme Gardou 708dc6b7d6 [NTOS:MM] Do not zero-fill pages when not needed 2021-08-04 17:48:39 +02:00
Jérôme Gardou 34662e2f6f [NTOS:MM] Copy the WorkingSet page when initializing "hand-built" process
CORE-17642
2021-08-04 11:49:38 +02:00
Jérôme Gardou be6a6b806a [NTOS:MM] Remove dead code 2021-08-03 10:57:12 +02:00
Jérôme Gardou b0dfe20981 [NTOS:MM] Fix MiIsUserPte inline 2021-07-30 18:54:54 +02:00
Jérôme Gardou ce65553eaa [NTOS:MM] Annotate some functions with regards to MmPfnLock
Also fix MI_ASSERT_PFN_LOCK_HELD macro.
Now in debug builds, SpinLocks are always non-zero when held.
2021-07-27 15:36:55 +02:00
Jérôme Gardou 6150dc994a [NTOS:MM] Always release PFN lock in MmDereferenceSegmentWithLock
All callers release it after calling it.
CORE-17698
2021-07-27 15:36:55 +02:00
Timo Kreuzer 47372e2ac9 [NTOS:MM] Fix MiSetSystemCodeProtection
CORE-16387 CORE-17638
Windows only makes the header RO for session space mapped drivers (win32k, dx and display drivers)
2021-07-11 21:43:07 +02:00
Jérôme Gardou 91587a432b [NTOS:MM] Fix another instance of reentrant spinlock acquisition 2021-06-29 11:49:20 +02:00
Jérôme Gardou 376fbc5242 [NTOS:MM] Check result of RtlCreateUnicodeString in MmGetFileNameForAddress
CORE-17637
2021-06-28 10:20:57 +02:00
Jérôme Gardou 59cddd15e2 [NTOS:MM] Implement MmFlushImageSection(MmFlushForDelete)
CORE-17544
2021-06-25 10:28:51 +02:00
Jérôme Gardou b3e9c89725 [NTOS:MM] Simplify refcounting when adding/deleting a page in a shared segment
CORE-17544
2021-06-25 10:28:51 +02:00
Jérôme Gardou bec42b6530 [NTOS:MM] Do not call ExAllocatePool* with PFN lock acquired
Unsurprisingly, it actually might acquire it
2021-06-25 10:28:51 +02:00
Jérôme Gardou c6e9fea844 [NTOSKRNL] Deduplicate some symbols 2021-06-24 18:48:31 +02:00
Jérôme Gardou 1c5ce8d11c [NTOS:MM] Fix-up call site of traced PFNs when serving a page fault 2021-06-24 18:48:31 +02:00
Jérôme Gardou 146337cc50 [NTOS:MM] Save caller address when tracing PFNs 2021-06-24 18:48:31 +02:00
Jérôme Gardou c8ce0cc434 [NTOS:MM] Fix one occurrence of MiDeletePte instead of MiDeletePde 2021-06-22 12:05:20 +02:00
Serge Gautherie eab2bb4b20 [NTOS:MM] MmFlushSegment(): Fix Iosb annotation
Addendum to 90c6a65.
2021-06-22 10:33:45 +02:00
Timo Kreuzer 396167bc7e [NTOS:MM] Improve an ASSERT 2021-06-19 12:17:18 +02:00
Timo Kreuzer 33949028fb [NTOS:MM] Fix PFN lock usage 2021-06-19 12:17:18 +02:00
Victor Perevertkin 3f601122bc [NTOS:MM] Fix placing entries into MmNonPagedPoolFreeListHead
When freeing pages, free page entries with pages num == 3 were
incorrectly treated as entries with pages num >= 4 and thus
their re-insertion was not triggered. That lead to non paged pool
fragmentation (can be triggered by kmtest:ExPools, for example)

Also, altered the index acquisition code for MmNonPagedPoolFreeList
entries so it looks more clear
2021-06-16 12:14:27 +02:00
Victor Perevertkin 34593d933b
[FORMATTING][NTOS][HAL][DRIVERS][SDK] Fix trailing spaces 2021-06-11 15:33:08 +03:00
Jérôme Gardou afc27ab1aa [NTOS:MM] Handle PT references for page table faults in MmAccessFault
Idea from Timo Kreuzer
2021-06-09 11:27:18 +02:00
Jérôme Gardou c7e09061ca [NTOS:MM] Get rid of MiQueryPageTableReferences 2021-06-09 11:27:18 +02:00
Timo Kreuzer 6a2eeaa5ae [NTOS] Implement MiIsPageTablePresent as a replacement for the abused MiQueryPageTableReferences 2021-06-09 11:27:18 +02:00
Jérôme Gardou 678923bf4b [NTOS:MM] In MiMakePdeExistAndValid, do not overly check PDE validity
It's "okay" to not #ifdef, but it's useless and confusing.
2021-06-09 11:27:18 +02:00
Jérôme Gardou c8fb3f7514 [NTOS:MM] Implement proper refcounting of page tables on amd64
CORE-17552
2021-06-09 11:27:18 +02:00
Jérôme Gardou 3aa346c21f [NTOS:MM] Make MiIsEntireRangeCommitted 64-bits aware 2021-06-09 11:27:18 +02:00
Jérôme Gardou 8bfba59b3e [NTOS:MM] Add some DPRINTs when incrementing or decrementing PFN share count 2021-06-09 11:27:18 +02:00
Jérôme Gardou a80ae8f2ea [NTOS:MM] Make MiCalculatePageCommitment 64-bits aware 2021-06-09 11:27:18 +02:00
Jérôme Gardou ed1038ff9c [NTOS:MM] Properly initialize initial process address space on amd64
By applying the same kludge on the relevant MMPFNs than on x86
2021-06-09 11:27:18 +02:00
Jérôme Gardou f9ea58dc97 [NTOS:MM] Unconditionally delete the PFN in MiDeletePte when PTE is in transition
When we will have a modified page writer, it will have to know if the written
page is stale when being done, and act accordingly.

CORE-8552
2021-06-09 11:27:18 +02:00
Jérôme Gardou 82dd1afe33 [NTOS:MM] Remove some dead code from legacy Mm 2021-06-09 11:27:18 +02:00
Jérôme Gardou a5ea259b28 [NTOS:MM] Fix remnants of CORE-17587 2021-06-07 17:04:18 +02:00
Jérôme Gardou 4ddc66f224 [NTOS:MM] Make MI_MAKE_HARDWARE_PTE_KERNEL actually respect the protection it is given
This one would set a page as writable & dirty even if MM_READONLY was passed
CORE-17587
2021-06-07 17:04:18 +02:00
Hervé Poussineau 831cad3410 [NTOS:MM] Remove wrong cleaning of PDEs at session cleanup 2021-05-25 21:13:32 +02:00
Hervé Poussineau 98882fb089 [NTOS:MM] Finish implementation of MmSessionDelete 2021-05-25 18:48:16 +02:00
Hervé Poussineau eac0afc6fe [NTOS:MM] MiDereferenceSession: return as soon as session is destroyed
This leads to some simplifications in the following code.
2021-05-25 18:48:16 +02:00
Jérôme Gardou 7eff8a36d5 Revert "[NTOS:MM] Add private pages to process working sets"
This is so full of bugs, I don't know what to say.
This reverts commit 374fef2d59.
2021-05-20 00:19:43 +02:00
Jérôme Gardou 969e60a06d [NTOS:MM] Move software PTE specific check after it's been tested for being one
Caught by The_DarkFire, only soul on earth actually *running* our SMP kernel
2021-05-19 23:56:21 +02:00
Jérôme Gardou 7d0d14965e [NTOS:MM] Do not map Process Working Set List in kernel space
This is not needed anymore and causes havoc on amd64
2021-05-19 23:17:27 +02:00
Jérôme Gardou 826aec48de [NTOS:MM] Fix build on whatever MSVC version triggered this
There is no real need to put static or extern "C" function in a special namespace anyway
2021-05-19 00:12:05 +02:00
Jérôme Gardou 374fef2d59 [NTOS:MM] Add private pages to process working sets 2021-05-18 23:42:19 +02:00
Jérôme Gardou 6a5fd8f487 [NTOS:MM] Relax requirements with regards to PFN lock when adding & removing entries in Working Sets
Once a page is used, holding the WS lock is enough until you want to free it.
2021-05-18 23:42:19 +02:00
Jérôme Gardou 3adf450867 [REACTOS] Addendum to 5c7ce4475e - Fix MSVC 2015 build
Put data into PAGEDATA or INITDATA sections
Keep section declaration for prototypes

CORE-17540
2021-05-12 11:04:29 +02:00
Jérôme Gardou f421bccbcc [NTOS:MM] First shot for Working Set list support
- Initialize
- Add private page (no shared page support yet)
- Remove pages
- Trim

Yes, this is C++ in the kernel.
2021-05-04 12:02:41 +02:00
Jérôme Gardou 31afbf1054 [NTOS] Make some internal headers C++ aware 2021-05-04 12:02:41 +02:00
Jérôme Gardou 8df48f9876 [NTOS:MM] Introduce MI_IS_PROCESS_WORKING_SET helper function 2021-05-04 12:02:41 +02:00
Jérôme Gardou 6b2f05f9dd [NTOS:MM] Implement turning working set shared lock to exclusive 2021-05-04 12:02:41 +02:00
Jérôme Gardou cd085ac12f [NTOS/MM] Implement Mi(Un)lockWorkingSetShared
Also fix checks when exclusively locking
2021-05-04 12:02:41 +02:00
Serge Gautherie 710acab83d [NTOS:MM] MiSetPagingOfDriver(): Add an explicit #if around unreachable code 2021-04-26 10:27:20 +02:00
Jérôme Gardou c48580135d [NTOS:MM] Fix a bit page fault handler with regard to COW sections 2021-04-08 15:40:37 +02:00
Jérôme Gardou a34d9bcfb6 [NTOS:MM] Share "page.c" between i386 & amd64 builds 2021-04-08 15:40:37 +02:00
Jérôme Gardou 7ea8312617 [NTOS:MM] Split MmCreateProcessAddressSpace in two parts
Generic one and arch-specific one.
Properly fail if we are out of resources.
Restore a lost assert.
2021-04-08 15:40:37 +02:00
Jérôme Gardou 6c027d28f9 [NTOS:MM] Properly track system process hyper space PDE page 2021-04-08 15:40:37 +02:00
Jérôme Gardou 370982aa1d [NTOS:MM] Make i386/page.c usable for all x86 variants
This includes PAE & amd64
2021-04-08 15:40:37 +02:00
Jérôme Gardou b10d92a16c [NTOS:MM] Use MI_MAKE_HARDWARE_PTE & friends in legacy Mm 2021-04-08 15:40:37 +02:00
Jérôme Gardou 43378411fb [NTOS:MM] Rewrite arch-specifics of the legacy Mm
Properly handle PDE refcounting
Clean-up of the internal API
Enforce attaching to the process when modifying its memory layout, instead of
making circonvoluted mappings which always end up being broken.
2021-04-08 15:40:37 +02:00
Jérôme Gardou b445005c70 [NTOS:MM] Get rid of MmRosFlushVirtualMemory
It's not used anywhere now, and it will eventually be gone
2021-04-08 15:40:37 +02:00
Jérôme Gardou 9e121fb6c2 [NTOS:MM] Get rid of MmSetCleanAllRmaps and MmIsDirtyPageRmap
Everything is wrong with them. Bad locking. Bad logic.
2021-04-08 15:40:37 +02:00
Jérôme Gardou 7bffb92099 [NTOS:MM] Let ARM3 handle invalid PDE for itself. 2021-04-08 15:40:37 +02:00
Jérôme Gardou 36a92e6ea5 [NTOS:MM] Fix a bit the page-out/page-in logic
- Do not lock the section segment when we are serving a fault for a process private page.
 - Do not keep the process address space lock while writing to pagefile.
 - Do not wait for an event that might never be set.
2021-04-08 15:40:37 +02:00
Jérôme Gardou 82c908195c Revert "[NTOS:MM] Allow MiMapPageInHyperSpace to be called from DISPATCH_LEVEL"
This reverts commit 8404d1a6ff.
Not ready for prime time, sorry.
2021-04-07 23:26:44 +02:00
Jérôme Gardou 8404d1a6ff [NTOS:MM] Allow MiMapPageInHyperSpace to be called from DISPATCH_LEVEL
Also annotate it to show its behaviour.
2021-04-07 23:09:26 +02:00
Oleg Dubinskiy 52f4d8019b [NTOS:MM][NTOS:SE] Mute some noisy stubs
Mute debug prints of MmDoesFileHaveUserWritableReferences and SeAuditingFileEventsWithContext stubs.
These stubs are very noisy and create a lot of spam in the log when using Microsoft NTFS driver in ReactOS (with some other improvements applied).
Implementing those functions isn't badly required for the proper work of this driver, so better way for now is just mute these stubs a bit.
After my changes, they will be displayed only once, and the log will be more clear, so it will seem to be enough to understand that the driver calls these routines.
CORE-17409
2021-04-07 13:16:03 +02:00
Mark Jansen 5e928e5c92
[NTOS:MM] Stubplement cookie generation for drivers 2021-04-03 18:08:53 +02:00
Jérôme Gardou 998870c5ea [NTOS:MM] Properly fail for invalid sizes of data section mappings 2021-03-30 22:20:15 +02:00
Jérôme Gardou 4c731adc04 [NTOS:MM] Fix compilation when PFN tracing is enabled 2021-03-30 17:22:09 +02:00
Jérôme Gardou 80f8beeeee [NTOS/MM] Remove an outdated check 2021-03-30 17:13:19 +02:00
Jérôme Gardou 0187c1e113 [NTOS:MM] Fix PFN tracing 2021-03-30 16:26:43 +02:00
Jérôme Gardou d1d1260f5f [NTOS:MM] Fix integer arithmetics 2021-03-22 10:02:12 +01:00
Jérôme Gardou 42094071ee [NTOS:MM] Silence noisy DPRINTs 2021-03-18 12:24:21 +01:00
Jérôme Gardou 4ff5d39edb [NTOS:MM] Use a define for the big pool table occupation rate 2021-03-18 12:24:21 +01:00
Jérôme Gardou 8494688eeb [NTOS:MM] Do not use atomic operations to set a bit while holding a lock.
Saving your fingers from typing it doesn't make it better code.
2021-03-18 12:24:21 +01:00
Jérôme Gardou f06b58925d [NTOS:MM] Implement shrinking big pool allocation table
Shrink when using 1/8 of its allocated capacity (thus use 25% of it at the end of the process)
Expand when using 3/4 of its allocated capacity (thus use ~40% of it at the end of the process)
2021-03-18 12:24:21 +01:00
Jérôme Gardou 54354712e2 [NTOS:MM] Fail allocating non-paged pool when running out of available pages 2021-03-18 12:24:21 +01:00
Jérôme Gardou 34630a06ca [NTOS:MM] Do not bugcheck when we are freeing a big allocation for which we didn't manage to insert the tag 2021-03-18 12:24:21 +01:00
Jérôme Gardou d5fa0788d8 [NTOS:MM] Silence a DPRINT which became noisy 2021-03-16 13:23:21 +01:00
Jérôme Gardou a93e065ebc [NTOS:MM] Do not chunk-read file when we are the modified-page-writer thread 2021-02-26 08:46:53 +01:00
Jérôme Gardou 35c091adaf [NTOS:MM] Do not chunk-read files marked with FO_RANDOM_ACCESS 2021-02-26 08:46:53 +01:00
Timo Kreuzer 4b4bff0642 [NTOS:MM] Handle demand zero PTEs with MM_EXECUTE_READWRITE and MM_READWRITE 2021-02-16 18:02:57 +01:00
Jérôme Gardou 0573de22a0 [NTOS:MM] Do not assert on a case we actually support 2021-02-10 17:48:30 +01:00
Jérôme Gardou 73fd52a164 [NTOS:MM] Flush dirty pages after closing the page files.
CORE-17462
2021-02-10 11:14:19 +01:00
Victor Perevertkin b705df731e
[NTOS:MM][FASTFAT_NEW] Fix DBG usage, fixes Release build 2021-02-05 11:54:10 +03:00
Jérôme Gardou 6f2b94c023 [NTOS:MM] Wake up the balancer as soon as we are low on page.
Fixes installing on a system with 64MB of RAM, if you feel like it.
2021-02-03 13:42:39 +01:00
Jérôme Gardou b7eb0fddf3 Address PR review 2021-02-03 09:41:24 +01:00
Jérôme Gardou c0bf352069 [NTOS:MM] Fix regression of MmCanFileBeTruncated 2021-02-03 09:41:24 +01:00
Jérôme Gardou 7f7abc9838 [NTOS:MM] Simplify code & fix MSVC x64 build 2021-02-03 09:41:23 +01:00
Jérôme Gardou 2ba1926037 [NTOS:MM][NTOS:CC] Performance improvement again
Read files by 64kb chunks instead of page-sized chunks.
2021-02-03 09:41:23 +01:00
Jérôme Gardou 625f273361 [NTOS:MM] More fine-tuning of the memory balancer 2021-02-03 09:41:23 +01:00