kib (Konstantin Belousov)
User

Projects

User Details

User Since
May 16 2014, 7:35 PM (179 w, 1 d)

Recent Activity

Today

kib added inline comments to D12753: kernel macro cleanliness, first pass.
Sun, Oct 22, 9:10 AM
kib committed rS324855: MFC r323772, r324302-r324308, r324310, r324313, r324315, r324326, r324330,.
MFC r323772, r324302-r324308, r324310, r324313, r324315, r324326, r324330,
Sun, Oct 22, 8:47 AM
kib committed rS324853: Remove the support for mknod(S_IFMT), which created dummy vnodes with.
Remove the support for mknod(S_IFMT), which created dummy vnodes with
Sun, Oct 22, 8:12 AM

Yesterday

kib committed rS324824: Check that the page which is freed as zeroed, indeed has all-zero content..
Check that the page which is freed as zeroed, indeed has all-zero content.
Sat, Oct 21, 5:28 PM

Fri, Oct 20

kib accepted D12745: Add OID for vm.overcommit..
Fri, Oct 20, 6:48 PM
kib committed rS324807: Take the vm object lock in read mode in vnode_generic_putpages()..
Take the vm object lock in read mode in vnode_generic_putpages().
Fri, Oct 20, 6:40 PM
kib accepted D12741: Don't call realpath(3) from libmap rtld code..
Fri, Oct 20, 4:15 PM
kib added inline comments to D12741: Don't call realpath(3) from libmap rtld code..
Fri, Oct 20, 3:12 PM
kib added inline comments to D12743: Make elf_aux_info() as public libc function..
Fri, Oct 20, 2:15 PM
kib accepted D12743: Make elf_aux_info() as public libc function..

This looks mostly fine, but is not committable because AT_HWCAP2 is not defined.

Fri, Oct 20, 2:10 PM
kib added inline comments to D12741: Don't call realpath(3) from libmap rtld code..
Fri, Oct 20, 1:09 PM
kib added a comment to D12725: Eliminate redundant TLB invalidations in the arm64 pmap.

I think the architecture allows us to issue the tlb invalidation, then later execute the dsb & isb instructions as long as we don't try to the old mapping, however will need to check with ARM. We should also just invalidate all entries in pmap_invalidate_range if the length is too long, however in this case it may be better to stop invalidating until after we have removed all the pagetable entries.

Fri, Oct 20, 10:13 AM
kib committed rS324796: MFC r324600, r324716:.
MFC r324600, r324716:
Fri, Oct 20, 9:21 AM
kib committed rS324795: Move swapout code into vm/vm_swapout.c..
Move swapout code into vm/vm_swapout.c.
Fri, Oct 20, 9:11 AM
kib closed D12663: Move swapout code into vm_swapout.c..
Fri, Oct 20, 9:10 AM
kib committed rS324794: Do not overwrite clean blocks on pageout..
Do not overwrite clean blocks on pageout.
Fri, Oct 20, 8:32 AM
kib closed D12668: Do not overwrite clean blocks on pageout..
Fri, Oct 20, 8:32 AM
kib committed rS324793: In vm_page_free_phys_pglist(), do not take vm_page_queue_free_mtx if.
In vm_page_free_phys_pglist(), do not take vm_page_queue_free_mtx if
Fri, Oct 20, 8:26 AM

Thu, Oct 19

kib added a comment to D12725: Eliminate redundant TLB invalidations in the arm64 pmap.
In D12725#264164, @alc wrote:
In D12725#264159, @kib wrote:

I agree that the change is functionally correct, but perhaps the intent was to coalesce the invalidations ? I.e. either pmap_remove_l3() should stop doing the invalidation, or its variant is added which does not do the invalidation and used there.

Then, for correctness, you have to introduce the delayed invalidation machinery from amd64. (Or perform a TLB range invalidation operation whenever you change PV list locks of the addresses whose PTEs were destroyed since the last PV list lock change.)

Thu, Oct 19, 6:30 PM
kib accepted D12730: Avoid the nbp lookup in the final loop iteration in flushbuflist().
Thu, Oct 19, 5:24 PM
kib added a comment to D12725: Eliminate redundant TLB invalidations in the arm64 pmap.

I agree that the change is functionally correct, but perhaps the intent was to coalesce the invalidations ? I.e. either pmap_remove_l3() should stop doing the invalidation, or its variant is added which does not do the invalidation and used there.

Thu, Oct 19, 5:19 PM

Wed, Oct 18

kib added a comment to D12699: Add AT_HWCAP2 ELF auxiliary vector. .

Perhaps split addition of AT_HWCAP2 and libc export commits.

Wed, Oct 18, 4:02 PM
kib added a comment to D12699: Add AT_HWCAP2 ELF auxiliary vector. .

My exp-run for selected ports just ended and It found unexpected problem.
Some ports detect getauxval() presence and if is present then expect that all (linux specific) AT_ flags are defined and implemented (e.g. security/p11-kit ).

I don't want to emulated full Linux behavior here, so it's time for plan B
As alternative we can make _elf_aux_info() public (as elf_aux_info()). If is this acceptable, can you please suggest me best/suitable header for this function?
Is (new) auxv.h still OK?

This approach result to slightly bigger patches for ports (that getauxval), but it's still acceptable I think.

Wed, Oct 18, 2:02 PM
kib added a comment to D12718: add libpmcstat.

libpmcstat.c must be split into per-function .c file. Also, I strongly suggest to mark the library as internal, since we cannot maintain PMC ABI even for a few weeks.

Wed, Oct 18, 1:58 PM
kib committed rS324718: MFC r324528:.
MFC r324528:
Wed, Oct 18, 8:06 AM
kib committed rS324716: Do not report reduction of swap zone if it was not..
Do not report reduction of swap zone if it was not.
Wed, Oct 18, 7:28 AM

Tue, Oct 17

kib updated the diff for D12668: Do not overwrite clean blocks on pageout..

Stop incrementing uio_offset.
Fix grammar.

Tue, Oct 17, 6:56 PM
kib accepted D12704: Fix a racy VI_DOOMED check in MNT_VNODE_FOREACH_ALL().
Tue, Oct 17, 6:11 PM
kib added a comment to D12638: mlx5(4) rx timestamps..

I'm fine with this as-is. I have not had time to use this as we originally intended (to modify LRO to track the timestamps of each coalesced packet). It may be several weeks before I can do this, so I'm fine with you committing it pending others approval.

Tue, Oct 17, 5:42 PM
kib added inline comments to D12699: Add AT_HWCAP2 ELF auxiliary vector. .
Tue, Oct 17, 3:54 PM
kib added a comment to D12638: mlx5(4) rx timestamps..

I patched this into our netflix tree. I confirmed that, even with rx timestamps enabled, the new functionality does not cause a significant performance loss in our case. There seems to be roughly one or two cache misses per mlx5e_poll_rx_cq() to check the p->priv->clbr_done > 2 condition, and roughly 8x as many for the timestamp handling itself. This pushes the cost of mlx5e_build_rx_mbuf() higher, but not horribly so.

See the attached vtune screencap.

Tue, Oct 17, 3:30 PM
kib added inline comments to D12699: Add AT_HWCAP2 ELF auxiliary vector. .
Tue, Oct 17, 3:22 PM
kib updated the diff for D12668: Do not overwrite clean blocks on pageout..

Change a condition into the assert. Add terminating dots into the comments.

Tue, Oct 17, 9:02 AM
kib added inline comments to D12668: Do not overwrite clean blocks on pageout..
Tue, Oct 17, 8:53 AM

Mon, Oct 16

kib committed rS324671: Re-evaluate thread' signal mask after ptracestop()..
Re-evaluate thread' signal mask after ptracestop().
Mon, Oct 16, 8:22 PM
kib committed rS324670: Improve assertion that an ignored or blocked signal is not delivered..
Improve assertion that an ignored or blocked signal is not delivered.
Mon, Oct 16, 8:15 PM
kib committed rS324669: Style..
Style.
Mon, Oct 16, 8:11 PM
kib committed rS324665: Fix the pv_chunks pc_lru tailq handling in reclaim_pv_chunk()..
Fix the pv_chunks pc_lru tailq handling in reclaim_pv_chunk().
Mon, Oct 16, 3:16 PM
kib updated the test plan for D12668: Do not overwrite clean blocks on pageout..
Mon, Oct 16, 1:42 PM

Sun, Oct 15

kib updated the diff for D12668: Do not overwrite clean blocks on pageout..

Update the patch to the current set of the bug fixed.

Sun, Oct 15, 7:37 PM
kib committed rS324624: MFC r324156:.
MFC r324156:
Sun, Oct 15, 10:59 AM

Sat, Oct 14

kib created D12668: Do not overwrite clean blocks on pageout..
Sat, Oct 14, 12:38 PM

Fri, Oct 13

kib updated the diff for D12663: Move swapout code into vm_swapout.c..

Remove empty line.
Do not ever return from NO_SWAPPING swapper().

Fri, Oct 13, 8:13 PM
kib accepted D12665: Reduce traffic on vm_cnt.v_free_count.
Fri, Oct 13, 8:09 PM
kib updated the summary of D12663: Move swapout code into vm_swapout.c..
Fri, Oct 13, 7:10 PM
kib created D12663: Move swapout code into vm_swapout.c..
Fri, Oct 13, 7:05 PM
kib committed rS324600: Evaluate the real size of the sblk_zone..
Evaluate the real size of the sblk_zone.
Fri, Oct 13, 4:23 PM
kib closed D12660: Evaluate the real size of the sblk_zone..
Fri, Oct 13, 4:23 PM
kib updated the diff for D12660: Evaluate the real size of the sblk_zone..

Add comment.

Fri, Oct 13, 3:50 PM
kib created D12660: Evaluate the real size of the sblk_zone..
Fri, Oct 13, 3:28 PM
kib accepted D12654: Support for Kabylake PMC.
Fri, Oct 13, 1:57 PM
kib added a comment to D12654: Support for Kabylake PMC.

Is there a single IAPDESCR in your patch which contains only one of two flags IAP_F_SL and IAP_F_KL ? I mean that SDM lists desktop Skylake and Kabylake core events in the same table.

Fri, Oct 13, 1:09 PM
kib added inline comments to D12650: Add EFI RTC support.
Fri, Oct 13, 9:47 AM

Wed, Oct 11

kib updated the diff for D12638: mlx5(4) rx timestamps..

hselasky notes.

Wed, Oct 11, 9:45 PM
kib added inline comments to D12638: mlx5(4) rx timestamps..
Wed, Oct 11, 9:43 PM
kib added a comment to D12638: mlx5(4) rx timestamps..
In D12638#262502, @imp wrote:

What's the quality of the free-running clock?

Wed, Oct 11, 6:46 PM
kib created D12638: mlx5(4) rx timestamps..
Wed, Oct 11, 4:04 PM
kib committed rS324528: The th_bintime, th_microtime and th_nanotime members of the timehand.
The th_bintime, th_microtime and th_nanotime members of the timehand
Wed, Oct 11, 11:03 AM

Mon, Oct 9

kib committed rS324439: Change amd64_get_ldt() to return 'EOF' when the LDT is not yet.
Change amd64_get_ldt() to return 'EOF' when the LDT is not yet
Mon, Oct 9, 4:20 PM
kib committed rS324438: Change i386_get_ldt() to return 'EOF' when the requested range of.
Change i386_get_ldt() to return 'EOF' when the requested range of
Mon, Oct 9, 4:19 PM
kib committed rS324437: Improvements to set_user_ldt()..
Improvements to set_user_ldt().
Mon, Oct 9, 4:07 PM
kib committed rS324433: Reset the fs and gs bases on exec(2)..
Reset the fs and gs bases on exec(2).
Mon, Oct 9, 3:39 PM
kib committed rS324432: More style..
More style.
Mon, Oct 9, 3:24 PM

Sun, Oct 8

kib committed rS324410: MFC r324301:.
MFC r324301:
Sun, Oct 8, 8:02 AM
kib committed rS324409: MFC r324174:.
MFC r324174:
Sun, Oct 8, 8:01 AM

Sat, Oct 7

kib committed rS324381: MFC r324114:.
MFC r324114:
Sat, Oct 7, 8:44 AM
kib committed rS324380: MFC r324113:.
MFC r324113:
Sat, Oct 7, 8:43 AM

Fri, Oct 6

kib accepted D12298: libc: remove gets.

I am fine with the libc bits of the patch.

Fri, Oct 6, 7:18 PM
kib added inline comments to D12298: libc: remove gets.
Fri, Oct 6, 3:24 PM
kib committed rS324366: Improve i386_get_ldt()..
Improve i386_get_ldt().
Fri, Oct 6, 2:30 PM
kib added a comment to D12298: libc: remove gets.

Can you put the readelf -a output from the patched libc somewhere ?

Fri, Oct 6, 12:52 PM
kib committed rS324355: Remove unneeded cast..
Remove unneeded cast.
Fri, Oct 6, 10:18 AM
kib committed rS324354: Style..
Style.
Fri, Oct 6, 10:17 AM

Thu, Oct 5

kib committed rS324334: Use ANSI C declarations..
Use ANSI C declarations.
Thu, Oct 5, 7:11 PM
kib committed rS324330: Correct format specifiers in the debug code. Style..
Correct format specifiers in the debug code. Style.
Thu, Oct 5, 6:58 PM
kib committed rS324326: Style..
Style.
Thu, Oct 5, 6:42 PM
kib accepted D12595: Add EFI RT support for arm64.
Thu, Oct 5, 1:17 PM
kib committed rS324315: Ensure that after sucessfull i386_set_ldt() call, other threads can.
Ensure that after sucessfull i386_set_ldt() call, other threads can
Thu, Oct 5, 1:13 PM
kib committed rS324313: Avoid a race betweem freeing LDT and context switches..
Avoid a race betweem freeing LDT and context switches.
Thu, Oct 5, 12:50 PM
kib committed rS324310: Improve amd64_get_ldt()..
Improve amd64_get_ldt().
Thu, Oct 5, 12:29 PM
kib committed rS324308: Minor style fix..
Minor style fix.
Thu, Oct 5, 12:20 PM
kib committed rS324307: Complete r323772 on amd64..
Complete r323772 on amd64.
Thu, Oct 5, 12:16 PM
kib committed rS324306: Use ANSI C declaration for amd64_get_ldt()..
Use ANSI C declaration for amd64_get_ldt().
Thu, Oct 5, 12:07 PM
kib committed rS324305: Correct format specifiers in the debug code..
Correct format specifiers in the debug code.
Thu, Oct 5, 12:02 PM
kib committed rS324304: Remove useless comments..
Remove useless comments.
Thu, Oct 5, 11:56 AM
kib committed rS324303: On amd64, mark the set_user_ldt() function as static..
On amd64, mark the set_user_ldt() function as static.
Thu, Oct 5, 11:50 AM
kib committed rS324302: Reduce default max_ldt_segment value to 512..
Reduce default max_ldt_segment value to 512.
Thu, Oct 5, 11:37 AM
kib committed rS324301: Update comment to note that we skip LDT reload for kthreads as well..
Update comment to note that we skip LDT reload for kthreads as well.
Thu, Oct 5, 11:35 AM
kib committed rS324300: MFC r324080:.
MFC r324080:
Thu, Oct 5, 11:01 AM
kib committed rS324299: MFC r324079:.
MFC r324079:
Thu, Oct 5, 11:00 AM
kib committed rS324298: MFC r323722,324078:.
MFC r323722,324078:
Thu, Oct 5, 10:47 AM

Wed, Oct 4

kib accepted D12589: vfs_lookup: Allow PATH_MAX-1 symlinks.
Wed, Oct 4, 4:51 PM
kib added inline comments to D12589: vfs_lookup: Allow PATH_MAX-1 symlinks.
Wed, Oct 4, 3:49 PM

Tue, Oct 3

kib accepted D12572: Various pathconf() fixes..
Tue, Oct 3, 11:12 AM
kib committed rS324235: MFC r323768:.
MFC r323768:
Tue, Oct 3, 11:00 AM

Mon, Oct 2

kib committed rS324191: Hide kernel stuff from userspace..
Hide kernel stuff from userspace.
Mon, Oct 2, 8:38 AM
kib accepted D12561: Restore the ability to deregister an eventhandler callback from within the callback..
Mon, Oct 2, 8:25 AM

Sun, Oct 1

kib committed rS324174: Fix supposed typo in the include guard symbol name, use full path for.
Fix supposed typo in the include guard symbol name, use full path for
Sun, Oct 1, 7:03 PM
kib accepted D12557: Add eventhandler notifications for newbus device attach/detach..

Note that in the current newbus locking model, DETACH_BEGIN and DETACH_COMPLETE/FAILED sequence of event handler invocation might interweave with other attach/detach events arbitrary. The handlers should be prepared for such situations.

Sun, Oct 1, 7:00 PM
kib committed rS324156: Improve smb(4) devfs interactions..
Improve smb(4) devfs interactions.
Sun, Oct 1, 11:17 AM