Page MenuHomeFreeBSD

kib (Konstantin Belousov)
User

Projects

User Details

User Since
May 16 2014, 7:35 PM (354 w, 6 d)

Recent Activity

Today

kib committed R10:e8b0f76b6130: O_RELATIVE_BENEATH: return ENOTCAPABLE instead of EINVAL for abs path (authored by kib).
O_RELATIVE_BENEATH: return ENOTCAPABLE instead of EINVAL for abs path
Thu, Mar 4, 7:28 PM
kib committed R10:af188c01b303: nameicap_check_dotdot: trim tracker on check (authored by kib).
nameicap_check_dotdot: trim tracker on check
Thu, Mar 4, 7:27 PM
kib committed R10:68230041eb29: Add nameicap_cleanup_from(), to clean tracker list starting from some element (authored by kib).
Add nameicap_cleanup_from(), to clean tracker list starting from some element
Thu, Mar 4, 7:27 PM
kib committed R10:11c6aa5c504d: nameicap_tracker_add: avoid duplicates in the tracker list (authored by kib).
nameicap_tracker_add: avoid duplicates in the tracker list
Thu, Mar 4, 7:27 PM
kib committed R10:7a1e21f194e4: Do not call nameicap_tracker_add() for dotdot case. (authored by kib).
Do not call nameicap_tracker_add() for dotdot case.
Thu, Mar 4, 7:26 PM
kib committed R10:6da53183f9ff: open(2): Remove O_BENEATH and AT_BENEATH (authored by kib).
open(2): Remove O_BENEATH and AT_BENEATH
Thu, Mar 4, 7:26 PM
kib committed R10:efc5f3d45469: fhlink(2): the syscalls do not take flag (authored by kib).
fhlink(2): the syscalls do not take flag
Thu, Mar 4, 7:26 PM
kib committed R10:def8b2b427ea: FFS extattr: fix handling of the tail (authored by kib).
FFS extattr: fix handling of the tail
Thu, Mar 4, 7:25 PM

Yesterday

kib added a comment to D29023: amd64: Only update fsbase/gsbase in pcb for curthread..
In D29023#650449, @jhb wrote:

If we added the assert we should by rights do it in all the cpu_fork/cpu_copy_thread impls.

Perhaps ok.

Wed, Mar 3, 10:54 PM
kib accepted D29049: link_elf_obj: Process init_array and fini_array.
Wed, Mar 3, 10:36 PM
kib added a comment to D29049: link_elf_obj: Process init_array and fini_array.

In principle it could be that an object contains both .ctors and SHT_INIT_ARRAY sections. I suggest to warn, at least.

Wed, Mar 3, 9:23 PM
kib committed R10:06eba90acb62: atomic: add atomic_interrupt_fence() (authored by kib).
atomic: add atomic_interrupt_fence()
Wed, Mar 3, 7:08 AM
kib accepted D28923: amd64 pmap: convert to counter(9), add PV and pagetable page counts.

I probably should have been more explicit, I proposed to drop the pt_page_count at all. But I hope that you will follow up with pmap_pt_page_alloc() patch so this does not matter much.

Wed, Mar 3, 5:51 AM
kib committed R10:e7be16de5044: atomic: add atomic_interrupt_fence() (authored by kib).
atomic: add atomic_interrupt_fence()
Wed, Mar 3, 3:50 AM
kib committed R10:c79e239ceeda: Add VOP_READ_PGCACHE(9) (authored by kib).
Add VOP_READ_PGCACHE(9)
Wed, Mar 3, 3:50 AM
kib added a comment to D29023: amd64: Only update fsbase/gsbase in pcb for curthread..

Should we explicitly set fsbase/gsbase for remote threads to some sane value, or assert that td1 != curthread always implies that td1 is kernel thread?

Wed, Mar 3, 3:36 AM
kib accepted D29026: amd64: Cleanups to setting TLS registers for Linux binaries..
Wed, Mar 3, 3:33 AM
kib added a comment to D29025: x86: Reset fsbase and gsbase for new threads..

But this change breaks ABI, I suspect. E.g. for amd64 current behavior is to set fsbase but gs/gsbase is copied. Similarly for 32bit on amd64 gsbase is set but fsbase is copied.

Wed, Mar 3, 3:31 AM
kib accepted D29024: x86: Don't bother copying the FPU state for cpu_copy_thread()..
Wed, Mar 3, 3:26 AM

Tue, Mar 2

kib added inline comments to D29016: capsicum-test: Update for O_BENEATH removal.
Tue, Mar 2, 6:53 PM
kib committed R10:28cd3a673e0e: O_RELATIVE_BENEATH: return ENOTCAPABLE instead of EINVAL for abs path (authored by kib).
O_RELATIVE_BENEATH: return ENOTCAPABLE instead of EINVAL for abs path
Tue, Mar 2, 6:24 PM
kib committed R10:49c98a4bf3a8: nameicap_check_dotdot: trim tracker on check (authored by kib).
nameicap_check_dotdot: trim tracker on check
Tue, Mar 2, 6:24 PM
kib committed R10:e8a2862aa038: Add nameicap_cleanup_from(), to clean tracker list starting from some element (authored by kib).
Add nameicap_cleanup_from(), to clean tracker list starting from some element
Tue, Mar 2, 6:23 PM
kib committed R10:2388ad7c293f: nameicap_tracker_add: avoid duplicates in the tracker list (authored by kib).
nameicap_tracker_add: avoid duplicates in the tracker list
Tue, Mar 2, 6:23 PM
kib committed R10:59e749428111: Do not call nameicap_tracker_add() for dotdot case. (authored by kib).
Do not call nameicap_tracker_add() for dotdot case.
Tue, Mar 2, 6:23 PM
kib committed R10:20e91ca36a56: open(2): Remove O_BENEATH and AT_BENEATH (authored by kib).
open(2): Remove O_BENEATH and AT_BENEATH
Tue, Mar 2, 6:23 PM
kib committed R10:600756afb532: fhlink(2): the syscalls do not take flag (authored by kib).
fhlink(2): the syscalls do not take flag
Tue, Mar 2, 6:23 PM
kib closed D28907: open(2): Remove O_BENEATH and AT_BENEATH.
Tue, Mar 2, 6:22 PM
kib committed R10:91a611e52160: mips: Don't set __NO_TLS to disable some uses of TLS. (authored by jhb).
mips: Don't set __NO_TLS to disable some uses of TLS.
Tue, Mar 2, 3:20 PM
kib committed R10:229f1e4b1424: riscv: Don't set __NO_TLS to disable some uses of TLS. (authored by jhb).
riscv: Don't set __NO_TLS to disable some uses of TLS.
Tue, Mar 2, 3:20 PM
kib committed R10:811deba9df26: Remove __NO_TLS. (authored by kib).
Remove __NO_TLS.
Tue, Mar 2, 3:20 PM
kib added a comment to D28923: amd64 pmap: convert to counter(9), add PV and pagetable page counts.

Can you split out pmap_alloc_pt_page() change from the other parts.

Tue, Mar 2, 11:03 AM
kib committed R10:49fe71bb1e8f: mips: Don't set __NO_TLS to disable some uses of TLS. (authored by jhb).
mips: Don't set __NO_TLS to disable some uses of TLS.
Tue, Mar 2, 12:50 AM
kib committed R10:d742dfbddd87: riscv: Don't set __NO_TLS to disable some uses of TLS. (authored by jhb).
riscv: Don't set __NO_TLS to disable some uses of TLS.
Tue, Mar 2, 12:50 AM
kib committed R10:73dc6674e08c: Remove __NO_TLS. (authored by kib).
Remove __NO_TLS.
Tue, Mar 2, 12:50 AM
kib committed R10:8742817ba62e: FFS extattr: fix handling of the tail (authored by kib).
FFS extattr: fix handling of the tail
Tue, Mar 2, 12:20 AM
kib closed D28999: FFS extattr: fix handling of the tail.
Tue, Mar 2, 12:20 AM

Mon, Mar 1

kib updated the test plan for D28999: FFS extattr: fix handling of the tail.
Mon, Mar 1, 3:36 PM
kib requested review of D28999: FFS extattr: fix handling of the tail.
Mon, Mar 1, 3:33 PM
kib updated subscribers of D28907: open(2): Remove O_BENEATH and AT_BENEATH.
Mon, Mar 1, 12:07 AM
kib updated the diff for D28907: open(2): Remove O_BENEATH and AT_BENEATH.

Grammar in man pages.
Change EINVAL to ENOTCAPABLE for abs path in O_RESOLVE_BENEATH.
Comment out O_UNUSED1/AT_UNUSED1.

Mon, Mar 1, 12:01 AM

Sun, Feb 28

kib closed D28980: Add VOP_READ_PGCACHE(9).
Sun, Feb 28, 11:39 PM
kib committed R10:55eb51ab6649: Add VOP_READ_PGCACHE(9) (authored by kib).
Add VOP_READ_PGCACHE(9)
Sun, Feb 28, 11:39 PM
kib accepted D28987: add safetybelt checks for cache_enter_time() when an NFS server sends a bogus reply.
Sun, Feb 28, 6:42 PM
kib added inline comments to D28980: Add VOP_READ_PGCACHE(9).
Sun, Feb 28, 6:38 PM
kib accepted D28974: filt_fsevent: only record interested events.
Sun, Feb 28, 5:25 PM
kib updated the diff for D28980: Add VOP_READ_PGCACHE(9).

Handle rwatson' notes.

Sun, Feb 28, 5:22 PM
kib added inline comments to D28980: Add VOP_READ_PGCACHE(9).
Sun, Feb 28, 5:19 PM
kib updated the diff for D28980: Add VOP_READ_PGCACHE(9).

Install. Cross-ref from vnode(9).

Sun, Feb 28, 1:15 AM
kib requested review of D28980: Add VOP_READ_PGCACHE(9).
Sun, Feb 28, 1:11 AM
kib committed R10:5d684701ca9a: Delete dead CLUSTERDEBUG config option. (authored by kib).
Delete dead CLUSTERDEBUG config option.
Sun, Feb 28, 12:44 AM
kib updated the diff for D28907: open(2): Remove O_BENEATH and AT_BENEATH.

Fix bug with removal of everything from tracker on a step.

Sun, Feb 28, 12:41 AM
kib added inline comments to D28907: open(2): Remove O_BENEATH and AT_BENEATH.
Sun, Feb 28, 12:39 AM

Sat, Feb 27

kib committed R10:b5449c92b489: Use atomic_interrupt_fence() instead of bare __compiler_membar() (authored by kib).
Use atomic_interrupt_fence() instead of bare __compiler_membar()
Sat, Feb 27, 11:29 PM
kib closed D28956: Use atomic_interrupt_fence() instead of bare __compiler_membar().
Sat, Feb 27, 11:28 PM
kib added inline comments to D28974: filt_fsevent: only record interested events.
Sat, Feb 27, 11:17 PM
kib added a comment to D28956: Use atomic_interrupt_fence() instead of bare __compiler_membar().

I'm not very familiar with bus_space_barrier(). I'm wondering if it's trying to handle the case where the resource is mapped WC?

Sat, Feb 27, 11:06 PM
kib added inline comments to D28923: amd64 pmap: convert to counter(9), add PV and pagetable page counts.
Sat, Feb 27, 10:55 PM
kib requested review of D28956: Use atomic_interrupt_fence() instead of bare __compiler_membar().
Sat, Feb 27, 12:05 AM

Fri, Feb 26

kib added inline comments to D28923: amd64 pmap: convert to counter(9), add PV and pagetable page counts.
Fri, Feb 26, 3:17 PM

Thu, Feb 25

kib added inline comments to D28923: amd64 pmap: convert to counter(9), add PV and pagetable page counts.
Thu, Feb 25, 11:31 PM
kib committed R10:f3a1daebaff5: Fix bug 253158 - Panic: snapacct_ufs2: bad block - mksnap_ffs(8) crash (authored by mckusick).
Fix bug 253158 - Panic: snapacct_ufs2: bad block - mksnap_ffs(8) crash
Thu, Feb 25, 8:56 PM
kib committed R10:6e3f19537ae6: Call softdep_prealloc() before taking ffs_lock_ea(), if unlock is committing (authored by kib).
Call softdep_prealloc() before taking ffs_lock_ea(), if unlock is committing
Thu, Feb 25, 8:55 PM
kib committed R10:b287d46c385b: ffs_close_ea: do not relock vnode under lock_ea (authored by kib).
ffs_close_ea: do not relock vnode under lock_ea
Thu, Feb 25, 8:55 PM
kib committed R10:16961e5e3fa4: ffs_vnops.c: style (authored by kib).
ffs_vnops.c: style
Thu, Feb 25, 8:55 PM
kib committed R10:9bef28f29446: ffs: do not call softdep_prealloc() from UFS_BALLOC() (authored by kib).
ffs: do not call softdep_prealloc() from UFS_BALLOC()
Thu, Feb 25, 8:55 PM
kib committed R10:298fde182f38: ffs_reallocblks: change the guard for softdep_prealloc() call to DOINGSUJ() (authored by kib).
ffs_reallocblks: change the guard for softdep_prealloc() call to DOINGSUJ()
Thu, Feb 25, 8:55 PM
kib committed R10:d1015ac54429: fifo: minor comment and assert improvements. (authored by kib).
fifo: minor comment and assert improvements.
Thu, Feb 25, 8:55 PM
kib committed R10:eed0fd5903e5: ffs_unlock: assert that IN_ENDOFF is not leaked past locked scope (authored by kib).
ffs_unlock: assert that IN_ENDOFF is not leaked past locked scope
Thu, Feb 25, 8:55 PM
kib committed R10:eadca59970c1: ffs softdep: Force processing of VI_OWEINACT vnodes when there is inode shortage (authored by kib).
ffs softdep: Force processing of VI_OWEINACT vnodes when there is inode shortage
Thu, Feb 25, 8:55 PM
kib committed R10:28d0541398bf: softdep_request_cleanup: wait for softdep_request_clean_flush() to pass (authored by kib).
softdep_request_cleanup: wait for softdep_request_clean_flush() to pass
Thu, Feb 25, 8:55 PM
kib committed R10:f8272ff1d26b: ufs_inactive(): stop hiding ERELOOKUP from ffs_truncate(), return it. (authored by kib).
ufs_inactive(): stop hiding ERELOOKUP from ffs_truncate(), return it.
Thu, Feb 25, 8:55 PM
kib committed R10:2e4940262c71: Stop ignoring ERELOOKUP from VOP_INACTIVE() (authored by kib).
Stop ignoring ERELOOKUP from VOP_INACTIVE()
Thu, Feb 25, 8:55 PM
kib committed R10:8ef69a7e8fdf: ufs vnops: brace softdep_prelink() with DOINGSUJ instead of DOINGSOFTDEP (authored by kib).
ufs vnops: brace softdep_prelink() with DOINGSUJ instead of DOINGSOFTDEP
Thu, Feb 25, 8:54 PM
kib committed R10:f889959e0a35: ffs softdep: remove will_direnter argument of softdep_prelink() (authored by kib).
ffs softdep: remove will_direnter argument of softdep_prelink()
Thu, Feb 25, 8:54 PM
kib committed R10:bccb8dda0ee2: ufs_direnter: directory truncation does not need special case for rename (authored by kib).
ufs_direnter: directory truncation does not need special case for rename
Thu, Feb 25, 8:54 PM
kib committed R10:9fc9a50119fa: ufs_rename: use VOP_VPUT_PAIR and rely on directory sync/truncation there (authored by kib).
ufs_rename: use VOP_VPUT_PAIR and rely on directory sync/truncation there
Thu, Feb 25, 8:54 PM
kib committed R10:b62d17802dcf: ufs_direnter: move directory truncation to ffs_vput_pair(). (authored by kib).
ufs_direnter: move directory truncation to ffs_vput_pair().
Thu, Feb 25, 8:54 PM
kib committed R10:0c3cdbf4293a: ffs_vput_pair(): try harder to recover from the vnode reclaim (authored by kib).
ffs_vput_pair(): try harder to recover from the vnode reclaim
Thu, Feb 25, 8:54 PM
kib committed R10:b3162f7d998b: FFS: implement special VOP_VPUT_PAIR(). (authored by kib).
FFS: implement special VOP_VPUT_PAIR().
Thu, Feb 25, 8:54 PM
kib committed R10:2b2c2c7128ea: nfsserver: use VOP_VPUT_PAIR(). (authored by kib).
nfsserver: use VOP_VPUT_PAIR().
Thu, Feb 25, 8:54 PM
kib committed R10:56119250bbfe: ffs_snapshot: use VOP_VPUT_PAIR after VOP_CREATE. (authored by kib).
ffs_snapshot: use VOP_VPUT_PAIR after VOP_CREATE.
Thu, Feb 25, 8:54 PM
kib added inline comments to D28846: amd64: make bcmp in libc just call memcmp.
Thu, Feb 25, 8:54 PM
kib committed R10:02c163e49b03: Use VOP_VPUT_PAIR() for eligible VFS syscalls. (authored by kib).
Use VOP_VPUT_PAIR() for eligible VFS syscalls.
Thu, Feb 25, 8:54 PM
kib committed R10:b4cc32336d31: nullfs: provide special bypass for VOP_VPUT_PAIR (authored by kib).
nullfs: provide special bypass for VOP_VPUT_PAIR
Thu, Feb 25, 8:54 PM
kib committed R10:984c11f5aaa0: Add VOP_VPUT_PAIR() with trivial default implementation. (authored by kib).
Add VOP_VPUT_PAIR() with trivial default implementation.
Thu, Feb 25, 8:54 PM
kib committed R10:c2007457027d: vn_open(): If the vnode is reclaimed during open(2), do not return error. (authored by kib).
vn_open(): If the vnode is reclaimed during open(2), do not return error.
Thu, Feb 25, 8:54 PM
kib committed R10:9ab31d3d60ad: ufs_direnter/SU: unconditionally UFS_UPDATE inode when extending directory (authored by kib).
ufs_direnter/SU: unconditionally UFS_UPDATE inode when extending directory
Thu, Feb 25, 8:53 PM
kib committed R10:57582311c3e0: ffs_syncvnode: only clear IN_NEEDSYNC after successfull sync (authored by kib).
ffs_syncvnode: only clear IN_NEEDSYNC after successfull sync
Thu, Feb 25, 8:53 PM
kib committed R10:6a61e5e0a1ed: Merge ufs_fhtovp() into ffs_inotovp(). (authored by kib).
Merge ufs_fhtovp() into ffs_inotovp().
Thu, Feb 25, 8:53 PM
kib committed R10:fc54784b47b6: ffs_inotovp(): interface to convert (ino, gen) into alive vnode (authored by kib).
ffs_inotovp(): interface to convert (ino, gen) into alive vnode
Thu, Feb 25, 8:53 PM
kib committed R10:8ecc621480bb: ffs: Add FFSV_REPLACE_DOOMED flag to ffs_vgetf() (authored by kib).
ffs: Add FFSV_REPLACE_DOOMED flag to ffs_vgetf()
Thu, Feb 25, 8:53 PM
kib committed R10:b95a45365676: ffs: call ufsdirhash_dirtrunc() right after setting directory size (authored by kib).
ffs: call ufsdirhash_dirtrunc() right after setting directory size
Thu, Feb 25, 8:53 PM
kib committed R10:041d9a1e057e: buf SU hooks: track buf_start() calls with B_IOSTARTED flag (authored by kib).
buf SU hooks: track buf_start() calls with B_IOSTARTED flag
Thu, Feb 25, 8:52 PM
kib accepted D28926: Close races in vm object chain traversal for unlock.
Thu, Feb 25, 7:14 PM
kib updated the diff for D28907: open(2): Remove O_BENEATH and AT_BENEATH.

Remove mentions of 'topping directory'. Reformulate O_RESOLVE_BENEATH based on Mark' suggestion, but make it shorter.

Thu, Feb 25, 6:13 PM
kib accepted D28922: pmap: Fix largemap restart checks in the kernel_maps sysctl handler.
Thu, Feb 25, 3:25 PM
kib added a comment to D28922: pmap: Fix largemap restart checks in the kernel_maps sysctl handler.

I do not quite understand the change.

Thu, Feb 25, 2:52 PM
kib added a comment to D28907: open(2): Remove O_BENEATH and AT_BENEATH.
In D28907#647543, @kib wrote:

I wonder if we should just treat O_BENEATH/AT_BENEATH as a (deprecated) alias for the new O_RESOLVE_BENEATH/AT_RESOLVE_BENEATH? I doubt there's any code that depended on being able to leave the start dir with O_BENEATH?

I would agree with the proposal, but I want to save bits in open(2) flags. It is too wasteful to leave a bit for legacy purposes, which formally should be never needed if we remove O_BENEATH before 13.0.

Sorry about that, for some reason I though O_BENEATH already existed in 12. I agree that completely removing it makes more sense if it was never available in a real release.

Thu, Feb 25, 2:07 PM
kib added a comment to D28907: open(2): Remove O_BENEATH and AT_BENEATH.

I wonder if we should just treat O_BENEATH/AT_BENEATH as a (deprecated) alias for the new O_RESOLVE_BENEATH/AT_RESOLVE_BENEATH? I doubt there's any code that depended on being able to leave the start dir with O_BENEATH?

Thu, Feb 25, 1:26 PM