kib (Konstantin Belousov)
User

Projects

User Details

User Since
May 16 2014, 7:35 PM (196 w, 6 d)

Recent Activity

Today

kib committed rS329864: Do not return out of bound pointers from intr_lookup_source()..
Do not return out of bound pointers from intr_lookup_source().
Fri, Feb 23, 11:21 AM
kib closed D14470: Do not return out of bound pointers from intr_lookup_source()..
Fri, Feb 23, 11:21 AM
kib committed rS329863: MFC r329416:.
MFC r329416:
Fri, Feb 23, 11:17 AM

Yesterday

kib created D14470: Do not return out of bound pointers from intr_lookup_source()..
Thu, Feb 22, 12:12 PM

Wed, Feb 21

kib committed rS329707: MFC r329254:.
MFC r329254:
Wed, Feb 21, 11:31 AM
kib committed rS329706: MFC r329256:.
MFC r329256:
Wed, Feb 21, 11:29 AM
kib committed rS329705: MFC r329254:.
MFC r329254:
Wed, Feb 21, 11:28 AM
kib committed rS329704: MFC r329252:.
MFC r329252:
Wed, Feb 21, 11:27 AM

Tue, Feb 20

kib committed rS329638: MFC r329216:.
MFC r329216:
Tue, Feb 20, 10:35 AM
kib committed rS329637: MFC r329215:.
MFC r329215:
Tue, Feb 20, 10:34 AM

Mon, Feb 19

kib committed rS329600: Do not free(9) uninitialized pointer..
Do not free(9) uninitialized pointer.
Mon, Feb 19, 7:08 PM
kib committed rS329560: MFC r329347:.
MFC r329347:
Mon, Feb 19, 10:59 AM

Sun, Feb 18

kib updated the diff for D14384: VM_WAIT rework.

Drop domain free mutex.

Sun, Feb 18, 11:59 PM
kib updated the test plan for D14384: VM_WAIT rework.
Sun, Feb 18, 10:42 AM
kib accepted D14422: Use UMA_ZONE_NODUMP for the DMAR map entry zone.
Sun, Feb 18, 10:40 AM

Sat, Feb 17

kib added inline comments to D14384: VM_WAIT rework.
Sat, Feb 17, 10:42 PM
kib updated the diff for D14384: VM_WAIT rework.

Ensure that pointers are dereferenced only once, add a comment about it.
Remove no longer useful control vars.

Sat, Feb 17, 10:39 PM
kib committed rS329462: MFC r328083,328096,328116,328119,328120,328128,328135,328153,328157,.
MFC r328083,328096,328116,328119,328120,328128,328135,328153,328157,
Sat, Feb 17, 6:00 PM
kib updated the diff for D14384: VM_WAIT rework.

Switch to vm_wait(vm_object_t);.
Rewrite vm_wait_domain() using vm_wait_wdoms().
Remove pagedaemon_wait().

Sat, Feb 17, 12:41 AM

Fri, Feb 16

kib committed rS329416: Remove unused symbols..
Remove unused symbols.
Fri, Feb 16, 11:19 PM
kib updated the diff for D14384: VM_WAIT rework.

Use DOMAIN_SUBSET, improve vm_wait_for_obj() generality.

Fri, Feb 16, 1:54 PM
kib updated the summary of D14384: VM_WAIT rework.
Fri, Feb 16, 1:52 PM
kib added inline comments to D14384: VM_WAIT rework.
Fri, Feb 16, 1:49 PM
kib committed rS329365: Use local symbol for offset..
Use local symbol for offset.
Fri, Feb 16, 1:33 PM
kib committed rS329347: Note that on amd64 pmap_enter(psind = 1) works (since r321378)..
Note that on amd64 pmap_enter(psind = 1) works (since r321378).
Fri, Feb 16, 1:08 AM

Thu, Feb 15

kib accepted D14390: Fix a memory leak introduced in r328426..
Thu, Feb 15, 10:15 PM
kib added inline comments to D14376: Add vm_page_alloc_pages_after()..
Thu, Feb 15, 6:22 PM
kib created D14384: VM_WAIT rework.
Thu, Feb 15, 6:10 PM
kib added inline comments to D14376: Add vm_page_alloc_pages_after()..
Thu, Feb 15, 1:03 PM
kib accepted D14355: Remove redundant rounding logic from PMAP initialization.
Thu, Feb 15, 11:45 AM
kib accepted D14379: Fix getdirentries(2) under 32-bit compat..

Please do not put the diffs for generated files neither in review nor in the main commit.

Thu, Feb 15, 11:30 AM

Wed, Feb 14

kib added a reviewer for D14353: hwpmc on non-SMP issue: jhb.
Wed, Feb 14, 2:04 PM
kib added a comment to D14353: hwpmc on non-SMP issue.
In D14353#301078, @br wrote:
In D14353#300792, @kib wrote:

I do not think that your statement about modules always build with MAXCPU > 1 is true. Modules, if build properly, are build against opt_*.h set from the configured kernel, and this would provide the right values for SMP and related symbols.

In other words, I believe that your issue is a user error, building modules outside the kernel configuration set and then using it on KBI-incompatible kernel. Yes, options SMP changes KBI.

No, both of my kernel & hwpmc.ko are product of 'make buildkernel'. And I load hwpmc.ko to the corresponding kernel.
Take a look at amd64/include/param.h, I think KLD_MODULE macro is defined when we build modules, in result MAXCPU is 256 for modules even with non-SMP kernel.

Wed, Feb 14, 2:04 PM
kib accepted D14362: nanosleep(2): Fix bogus incrementing of rmtp by tc_tick_sbt on [EINTR]..
Wed, Feb 14, 1:59 PM
kib committed rS329256: Cleanup unused page argument for vm_reserv_break()..
Cleanup unused page argument for vm_reserv_break().
Wed, Feb 14, 12:34 AM
kib closed D14364: Cleanup unused page argument for vm_reserv_break()..
Wed, Feb 14, 12:34 AM
This revision was not accepted when it landed; it landed in state Needs Review.
Wed, Feb 14, 12:32 AM
kib committed rS329254: Ensure memory consistency on COW..
Ensure memory consistency on COW.
Wed, Feb 14, 12:32 AM
kib reopened D14347: Ensure memory consistency on COW.
Wed, Feb 14, 12:27 AM
kib closed D14347: Ensure memory consistency on COW.
Wed, Feb 14, 12:25 AM
kib committed rS329252: Do not call pmap_enter() with invalid protection mode..
Do not call pmap_enter() with invalid protection mode.
Wed, Feb 14, 12:25 AM

Tue, Feb 13

kib created D14364: Cleanup unused page argument for vm_reserv_break()..
Tue, Feb 13, 11:46 PM
kib accepted D14282: Set the RW/NX bits to protect kernel text and data loaded as part of the main kernel image.
Tue, Feb 13, 9:10 PM
kib added inline comments to D14355: Remove redundant rounding logic from PMAP initialization.
Tue, Feb 13, 9:09 PM
kib added a comment to D14353: hwpmc on non-SMP issue.

I do not think that your statement about modules always build with MAXCPU > 1 is true. Modules, if build properly, are build against opt_*.h set from the configured kernel, and this would provide the right values for SMP and related symbols.

Tue, Feb 13, 6:17 PM
kib committed rS329216: linuxkpi: Do not leak pages on put..
linuxkpi: Do not leak pages on put.
Tue, Feb 13, 3:44 PM
kib committed rS329215: Do not leak rv->psind in some specific situations..
Do not leak rv->psind in some specific situations.
Tue, Feb 13, 3:36 PM
kib closed D14335: Do not leak rv->psind in some weird situations..
Tue, Feb 13, 3:36 PM
kib committed rS329214: Fix build with gas..
Fix build with gas.
Tue, Feb 13, 3:30 PM
kib created D14347: Ensure memory consistency on COW.
Tue, Feb 13, 1:20 PM
kib committed rS329199: MFC r328913:.
MFC r328913:
Tue, Feb 13, 12:54 PM
kib added inline comments to D14282: Set the RW/NX bits to protect kernel text and data loaded as part of the main kernel image.
Tue, Feb 13, 8:13 AM

Mon, Feb 12

kib accepted D14282: Set the RW/NX bits to protect kernel text and data loaded as part of the main kernel image.

I am fine with this, but would be curious to see an effect of setting a breakpoint with ddb when write-protection is enabled for the kernel text.

Mon, Feb 12, 11:33 PM
kib added inline comments to D14282: Set the RW/NX bits to protect kernel text and data loaded as part of the main kernel image.
Mon, Feb 12, 10:39 PM
kib accepted D14333: Set the NX bit for the PML4 recursive mapping.
Mon, Feb 12, 10:15 PM
kib added a comment to D14335: Do not leak rv->psind in some weird situations..

BTW, I do not see any called for vm_reserv_break() which passes non-NULL m.

The last such caller went away in r308691, when PG_CACHED was removed.

Mon, Feb 12, 10:09 PM
kib created D14335: Do not leak rv->psind in some weird situations..
Mon, Feb 12, 7:48 PM
kib committed rS329156: MFC r328880:.
MFC r328880:
Mon, Feb 12, 10:34 AM

Sat, Feb 10

kib accepted D14263: read-behind / read-ahead support for zfs_getpages().
Sat, Feb 10, 4:43 PM

Fri, Feb 9

kib added inline comments to D14282: Set the RW/NX bits to protect kernel text and data loaded as part of the main kernel image.
Fri, Feb 9, 7:26 PM
kib added inline comments to D14282: Set the RW/NX bits to protect kernel text and data loaded as part of the main kernel image.
Fri, Feb 9, 6:22 PM
kib accepted D14269: Mark initial page table entries as wired.
Fri, Feb 9, 6:12 PM

Thu, Feb 8

kib added a comment to D14269: Mark initial page table entries as wired.

Is there any other reason than the consistency with other page table pages for this change ?

Thu, Feb 8, 6:10 PM
kib accepted D14268: Correctly mark as used physical blocks used to map the kernel.
Thu, Feb 8, 6:10 PM
kib accepted D14267: loader: fix endianness conversion.
Thu, Feb 8, 5:39 PM
kib accepted D14266: Use vm_page_unwire_noq() in some pmap code..
Thu, Feb 8, 5:34 PM
kib added inline comments to D14263: read-behind / read-ahead support for zfs_getpages().
Thu, Feb 8, 2:18 PM

Wed, Feb 7

kib accepted D13682: Add memset_s(3) doc to memset(3) man page..
Wed, Feb 7, 2:30 PM
kib added inline comments to D13682: Add memset_s(3) doc to memset(3) man page..
Wed, Feb 7, 10:25 AM

Tue, Feb 6

kib committed rS328961: Adjust MFC in r328944 to stable/11..
Adjust MFC in r328944 to stable/11.
Tue, Feb 6, 11:50 PM
kib added inline comments to D11943: Modify vm_page_wire() to not dequeue the specified page.
Tue, Feb 6, 8:00 PM
kib added inline comments to D14220: ld.lld.1: document the rest of the -z options.
Tue, Feb 6, 7:53 PM
kib added inline comments to D14220: ld.lld.1: document the rest of the -z options.
Tue, Feb 6, 7:02 PM
kib added inline comments to D11943: Modify vm_page_wire() to not dequeue the specified page.
Tue, Feb 6, 6:52 PM
kib added inline comments to D11943: Modify vm_page_wire() to not dequeue the specified page.
Tue, Feb 6, 6:50 PM
kib committed rS328913: Move signal trampolines out of locore.s into separate source file..
Move signal trampolines out of locore.s into separate source file.
Tue, Feb 6, 12:02 AM

Mon, Feb 5

kib added inline comments to D14212: Add descriptions to undocumented sysctls (hw, kern, vm).
Mon, Feb 5, 11:09 PM
kib accepted D13945: Always give ELF brands a chance to veto a match..
Mon, Feb 5, 6:20 PM
kib closed D14184: On munlock(), unwire correct page..
Mon, Feb 5, 12:49 PM
kib committed rS328880: On munlock(), unwire correct page..
On munlock(), unwire correct page.
Mon, Feb 5, 12:49 PM
kib added a comment to D13485: Improve pmap_free_zero_pages() operations.

I have only style notes about the patch. In my opinion, this is a good change if only due to the reduction of the code duplication.

Mon, Feb 5, 12:48 PM
kib committed rS328879: MFC r328773:.
MFC r328773:
Mon, Feb 5, 10:30 AM

Sat, Feb 3

kib updated the diff for D14184: On munlock(), unwire correct page..

Iterate over the locked_depth var.

Sat, Feb 3, 8:36 PM
kib created D14184: On munlock(), unwire correct page..
Sat, Feb 3, 12:13 PM

Fri, Feb 2

kib committed rS328773: On pageout, in vnode generic pager, for partially dirty page, only.
On pageout, in vnode generic pager, for partially dirty page, only
Fri, Feb 2, 11:56 AM

Wed, Jan 31

kib committed rS328628: When switching IBRS on, also enable STIBP (Single Thread Indirect.
When switching IBRS on, also enable STIBP (Single Thread Indirect
Wed, Jan 31, 4:56 PM
kib committed rS328627: Expand IBRS TLA in sysctl help lines..
Expand IBRS TLA in sysctl help lines.
Wed, Jan 31, 4:54 PM
kib committed rS328625: IBRS support, AKA Spectre hardware mitigation..
IBRS support, AKA Spectre hardware mitigation.
Wed, Jan 31, 2:36 PM
This revision was not accepted when it landed; it landed in state Needs Review.
Wed, Jan 31, 2:36 PM
kib committed rS328624: Do not enable PTI when IA32_ARCH_CAP_RDCL_NO bit is set..
Do not enable PTI when IA32_ARCH_CAP_RDCL_NO bit is set.
Wed, Jan 31, 2:25 PM
kib updated the diff for D14029: Spectre AKA IBRS.

Pre-calculate all IBRS pre-conditions instead of doing it on each kernel entry.
Disable IBRS on return if it was enabled, regardless of the IBRS enable state.

Wed, Jan 31, 11:46 AM

Tue, Jan 30

kib added inline comments to D14000: per-domain page queue free locking.
Tue, Jan 30, 7:02 PM
kib added inline comments to D14000: per-domain page queue free locking.
Tue, Jan 30, 6:48 PM
kib accepted D13459: kern.ipc.{msqids,semsegs,sema} sysctls for FreeBSD32..
Tue, Jan 30, 5:34 PM
kib added inline comments to D14000: per-domain page queue free locking.
Tue, Jan 30, 1:38 PM
kib added a comment to D14092: Make g_label depend on FFS for now..

For static compilation, this is perhaps fine.

Tue, Jan 30, 9:05 AM

Mon, Jan 29

kib added a comment to D13999: add map header sentinel to vm_map search tree.
In D13999#295780, @kib wrote:

Is there an assert somewhere that the map header is never returned by a lookup ?

Without this change, it is possible that vm_map_lookup_entry will return with *entry == &map->header, when the root is null or when a binary search is made for an address smaller than any in the tree. So there's no assertion to fail on that condition, and it would be incorrect to add one.

That will still be possible after this change, so that there's no cause to add an assertion.

Mon, Jan 29, 12:30 PM
kib committed rS328539: MFC r328264:.
MFC r328264:
Mon, Jan 29, 10:19 AM

Sun, Jan 28

kib added a comment to D13999: add map header sentinel to vm_map search tree.

Is there an assert somewhere that the map header is never returned by a lookup ?

Sun, Jan 28, 9:59 AM
kib accepted D13682: Add memset_s(3) doc to memset(3) man page..
Sun, Jan 28, 9:55 AM