mav (Alexander Motin)
User

Projects

User Details

User Since
Jun 3 2014, 6:27 PM (181 w, 1 d)

Recent Activity

Today

mav committed rS326120: MFC r325571: Add some PCI IDs found on AMD Epyc system..
MFC r325571: Add some PCI IDs found on AMD Epyc system.
Thu, Nov 23, 10:10 AM

Yesterday

mav committed rS326105: MFC r325552: s/NgSendMsgReply/NgSendReplyMsg/ in man to match the code..
MFC r325552: s/NgSendMsgReply/NgSendReplyMsg/ in man to match the code.
Wed, Nov 22, 9:24 PM
mav committed rS326104: MFC r325552: s/NgSendMsgReply/NgSendReplyMsg/ in man to match the code..
MFC r325552: s/NgSendMsgReply/NgSendReplyMsg/ in man to match the code.
Wed, Nov 22, 9:22 PM

Thu, Nov 9

mav committed rS325571: Add some PCI IDs found on AMD Epyc system..
Add some PCI IDs found on AMD Epyc system.
Thu, Nov 9, 10:16 AM

Wed, Nov 8

mav committed rS325552: s/NgSendMsgReply/NgSendReplyMsg/ in man to match the code..
s/NgSendMsgReply/NgSendReplyMsg/ in man to match the code.
Wed, Nov 8, 12:35 PM

Tue, Nov 7

mav added a comment to D12945: Introduce phys2disk..

I agree with above commenters that this is too narrow and specific. For example, for mps/mpr HBAs BTL addresses are tied not to physical location, but to disk SAS addresses. It means in case of disk replacement it will likely change, making any numbering based on top unstable.

Tue, Nov 7, 3:51 PM

Thu, Nov 2

mav committed rS325311: MFC r324752: Relax per-ifnet cif_vrs list double locking in carp(4)..
MFC r324752: Relax per-ifnet cif_vrs list double locking in carp(4).
Thu, Nov 2, 10:38 AM

Mon, Oct 30

mav accepted D12809: geom_slice: do not destroy softc until providers are gone.

Looks good to me. Thank you.

Mon, Oct 30, 7:49 AM

Sun, Oct 29

mav added a comment to D12809: geom_slice: do not destroy softc until providers are gone.
In D12809#266212, @avg wrote:
In D12809#266162, @mav wrote:

The idea looks fine, but logic of gsp->nopen and d_nopen in g_slice_access() looks over-engineered. I suppose gsp->nopen will never be above 1 here.

I think that gsp->nopen can be as high as slices parameter passed to g_slice_new that is greater than one for several geom classes.
There is some confusing naming, but gsp does not correspond to a single slice, it represents a "slicer" with potentially many slices.

Sun, Oct 29, 8:16 PM
mav added inline comments to D12797: extend illumos compatibility log_sysevent to support more nvlist types.
Sun, Oct 29, 11:17 AM
mav added a comment to D12809: geom_slice: do not destroy softc until providers are gone.

The idea looks fine, but logic of gsp->nopen and d_nopen in g_slice_access() looks over-engineered. I suppose gsp->nopen will never be above 1 here. I think you could even use just (cp->acr == 0 && cp->acw == 0 && cp->ace == 0) expression instead of additional open counter.

Sun, Oct 29, 10:42 AM
mav added a comment to D12809: geom_slice: do not destroy softc until providers are gone.
In D12809#266148, @avg wrote:

Just to clarify, do you mean adding tracking of the open provider count into g_slice_access ?

Sun, Oct 29, 9:07 AM
mav added a comment to D12809: geom_slice: do not destroy softc until providers are gone.

It would have been sufficient to wait until the providers leave the doorstep (upon nend reaching nstart), but there is no notification for that.

Sun, Oct 29, 8:49 AM

Oct 23 2017

mav accepted D12731: vdev_geom_close: close errored consumer even if vdev_reopening is set.

I have no objections.

Oct 23 2017, 9:56 AM
mav committed rS324909: MFC r324661: Add Creative vendor ID..
MFC r324661: Add Creative vendor ID.
Oct 23 2017, 7:42 AM
mav committed rS324908: MFC r324659: Update details of interface capabilities changed by bridge(4)..
MFC r324659: Update details of interface capabilities changed by bridge(4).
Oct 23 2017, 7:39 AM

Oct 19 2017

mav committed rS324752: Relax per-ifnet cif_vrs list double locking in carp(4)..
Relax per-ifnet cif_vrs list double locking in carp(4).
Oct 19 2017, 9:01 AM

Oct 16 2017

mav committed rS324661: Add Creative vendor ID..
Add Creative vendor ID.
Oct 16 2017, 12:55 PM
mav committed rS324659: Update details of interface capabilities changed by bridge(4)..
Update details of interface capabilities changed by bridge(4).
Oct 16 2017, 12:33 PM
mav accepted D12679: Document missing libgeom functions.
Oct 16 2017, 7:52 AM

Oct 14 2017

mav accepted D12652: iscsi: do not hold the global lock while tearing down a session.
Oct 14 2017, 9:09 AM
mav accepted D12653: no need to retry oustanding requests when terminating a session.

Looks good to me.

Oct 14 2017, 9:02 AM

Oct 8 2017

mav committed rS324408: MFC r324172: Align test I/O buffer to page boundary..
MFC r324172: Align test I/O buffer to page boundary.
Oct 8 2017, 7:20 AM
mav committed rS324407: MFC r324155: Add initial support for Address Lookup Table (A-LUT)..
MFC r324155: Add initial support for Address Lookup Table (A-LUT).
Oct 8 2017, 7:20 AM

Oct 7 2017

mav committed rS324382: MFC r324123: Add sysctl/tunable for maximal request time..
MFC r324123: Add sysctl/tunable for maximal request time.
Oct 7 2017, 9:47 AM

Oct 5 2017

mav committed rS324297: MFC r324086: Alike to ZFS disable cache flush after first ENOTSUP error..
MFC r324086: Alike to ZFS disable cache flush after first ENOTSUP error.
Oct 5 2017, 10:24 AM

Oct 4 2017

mav accepted D12578: Display TRIM/UNMAP support in "diskinfo -v".
Oct 4 2017, 9:40 AM

Oct 3 2017

mav added a comment to D12578: Display TRIM/UNMAP support in "diskinfo -v".

Generally looks good to me, but I'd probably moved it somewhere before zone mode, and may be also added then GEOM::rotation_rate, GEOM::lunid and GEOM::lunname for completeness.

Oct 3 2017, 7:55 PM

Oct 1 2017

mav committed rS324172: Align test I/O buffer to page boundary..
Align test I/O buffer to page boundary.
Oct 1 2017, 4:59 PM
mav committed rS324155: Add initial support for Address Lookup Table (A-LUT)..
Add initial support for Address Lookup Table (A-LUT).
Oct 1 2017, 9:49 AM

Sep 30 2017

mav committed rS324123: Add sysctl/tunable for maximal request time..
Add sysctl/tunable for maximal request time.
Sep 30 2017, 1:17 PM

Sep 28 2017

mav added a comment to D12531: Fix the error message when creating a zpool on a too-small device.

I am not sure think this problem has disappeared since this check was added. While it is less likely to see floppy drive these days, it is still not very reasonable to look for metadata on devices (and partitions) where they can not exist by definition (what IIRC Solaris never does unlike FreeBSD). I think as alternative solution to the reported problem size could be checked not on every open, but only in cases when metadata is going to be read. If on pool creation ZFS core requests specific device to be opened forcefully, we may just do that and let core reject it later with whatever error it prefers.

Sep 28 2017, 5:30 PM
mav committed rS324086: Alike to ZFS disable cache flush after first ENOTSUP error..
Alike to ZFS disable cache flush after first ENOTSUP error.
Sep 28 2017, 3:58 PM

Sep 22 2017

mav added inline comments to D12457: Complete INADDR_HASH lock protection.
Sep 22 2017, 12:18 PM
mav added a comment to D12086: ng_ether should notify lower and orphan nodes when underlying interface link layer address is changed.

PS: Read how to create patches with context for the phabricator. May be it is a problem of phabricator and it modified the diff after you, if I download the raw diff here, there is no even normal diff -p context, which is almost always useful.

Sep 22 2017, 11:58 AM · network
mav added a comment to D12086: ng_ether should notify lower and orphan nodes when underlying interface link layer address is changed.

While I understand your valid motivation, I can't say that I very much like this in patch for several reasons:

  • NGM_ETHER_SET_ENADDR is a control command supposed to be targeted to ng_ether node, not from it, and it is a action request, not notification. I suppose if you connect two ng_ether nodes with lower hooks to create poor mans bridge, this won't work as expected.
  • ng_pppoe has method to explicitly set MAC address, and it may get overwritten by this, which may or may not be correct.
  • We just fought one locking issue, and here I don't see any hook locks or even references (I understand that your problem is narrower, but still).
Sep 22 2017, 11:53 AM · network

Sep 21 2017

mav accepted D12435: Protect ng_iface(4) private data.

Looks good to me.

Sep 21 2017, 7:57 PM
mav committed rS323844: MFC r323575: Add second entry to LUT on a link side in B2B mode..
MFC r323575: Add second entry to LUT on a link side in B2B mode.
Sep 21 2017, 7:35 AM

Sep 20 2017

mav accepted D12228: Clear G_CF_ORPHAN when attaching..

I have no specific objections. Probably I was just thinking that consumer should be destroyed and recreated after orphan event happened, but if gmountver does not go it, this indeed may help. I just think whether G_CF_SPOILED should also be cleared there then?

Sep 20 2017, 9:20 PM

Sep 19 2017

mav added a comment to D12396: Default ABD chunk size.

Fragmentation is a separate issue compared to the outright waste incurred with a compressed ARC and small-enough record sizes.

Sep 19 2017, 10:42 PM · ZFS

Sep 18 2017

mav accepted D11779: Use nstosbt() instead of multiplying by SBT_1NS to avoid roundoff errors..

Looks good to me.

Sep 18 2017, 2:00 AM
mav added a comment to D12396: Default ABD chunk size.

I'll second @mahrens worry about possible memory fragmentation. Plus FreeBSD kernel memory allocator does not even have the mentioned hooks to reallocate memory fighting it. Plus I still hope we can make GEOM vdev to use unmapped I/O to avoid extra buffer copying, which depends on blocks to be multiple of page size (4KB). I am not against making it loader tunable, but not ready to change the default or promote the tunable.

Sep 18 2017, 1:46 AM · ZFS

Sep 14 2017

mav committed rS323575: Add second entry to LUT on a link side in B2B mode..
Add second entry to LUT on a link side in B2B mode.
Sep 14 2017, 4:51 AM

Sep 13 2017

mav added a comment to D12355: MFV r323535: 8585 improve batching done in zil_commit().

I red it quite briefly, but I suspect that use of cv_timedwait_sbt() together gethrtime() may not work right due to rounding issues. As is was already mentioned by others, in most cases use of SBT_1NS is a direct request for troubles. My fault.

Sep 13 2017, 4:58 PM

Sep 11 2017

mav committed rS323456: MFC r323128: Increase negotiation polling period from 10ms to 100ms..
MFC r323128: Increase negotiation polling period from 10ms to 100ms.
Sep 11 2017, 6:51 PM
mav committed rS323455: MFC r323126: Make NTB drivers report more info via NewBus methods..
MFC r323126: Make NTB drivers report more info via NewBus methods.
Sep 11 2017, 6:51 PM
mav committed rS323454: MFC r323074: Clear doorbell bits after masking them before processing..
MFC r323074: Clear doorbell bits after masking them before processing.
Sep 11 2017, 6:50 PM
mav committed rS323453: MFC r323032, r323053, r323058, r323059, r323084, r323114, r323127:.
MFC r323032, r323053, r323058, r323059, r323084, r323114, r323127:
Sep 11 2017, 6:48 PM
mav committed rS323452: MFC r322981: Mask doorbells while processing them..
MFC r322981: Mask doorbells while processing them.
Sep 11 2017, 6:32 PM
mav committed rS323451: MFC r322980: Fix fake interrupt when set doorbell is unmasked..
MFC r322980: Fix fake interrupt when set doorbell is unmasked.
Sep 11 2017, 6:31 PM

Sep 8 2017

mav accepted D12263: Remove spaces from CTL devices' default serial numbers.

I have no objections against the change. Though the whole concept of using so simple and repeatable LUN IDs is a request for troubles caused by conflicts.

Sep 8 2017, 2:33 PM

Sep 7 2017

mav added a comment to D12230: Make md(4) support GEOM::ident..

I don't see a point in reporting random ident strings. Those are normally useful to make things persistent, but random is not a way for that. May be I'd see it more useful to have it user-configurable.

Sep 7 2017, 3:57 PM
mav added a comment to D12130: Greatly reduce interrupt latency caused by steal_idle.

At high context switching rates on systems with many cores stealing code does create significant CPU load. So I can believe that the critical section can indeed be an issue, so this way may be good to go. I personally thought about limiting maximal stealing distance based on some statistic factors, such as context switch rates, etc, since we shouldn't probably touch every CPU caches on every context switch -- its a dead end with growing number of CPUs, but haven't got far on it.

Sep 7 2017, 3:50 PM
mav committed rS323265: MFC r323046: Make ntb_transport(4) ready receive early link events..
MFC r323046: Make ntb_transport(4) ready receive early link events.
Sep 7 2017, 12:51 PM
mav committed rS323264: MFC r323047: Make ntb_set_ctx() always generate fake link event..
MFC r323047: Make ntb_set_ctx() always generate fake link event.
Sep 7 2017, 12:50 PM

Sep 3 2017

mav committed rS323137: MFC r323045: Fix flags field decoding in ACPI_NFIT_CONTROL_REGION..
MFC r323045: Fix flags field decoding in ACPI_NFIT_CONTROL_REGION.
Sep 3 2017, 6:43 AM

Sep 2 2017

mav committed rS323128: Increase negotiation polling period from 10ms to 100ms..
Increase negotiation polling period from 10ms to 100ms.
Sep 2 2017, 1:29 PM
mav committed rS323127: Some NTB man pages polishing..
Some NTB man pages polishing.
Sep 2 2017, 11:57 AM
mav committed rS323126: Make NTB drivers report more info via NewBus methods..
Make NTB drivers report more info via NewBus methods.
Sep 2 2017, 11:56 AM

Sep 1 2017

mav committed rS323114: Sync NTB options between amd64 and i386..
Sync NTB options between amd64 and i386.
Sep 1 2017, 7:16 PM
mav committed rS323084: Link Interface has no Link Error registers..
Link Interface has no Link Error registers.
Sep 1 2017, 9:48 AM

Aug 31 2017

mav committed rS323074: Clear doorbell bits after masking them before processing..
Clear doorbell bits after masking them before processing.
Aug 31 2017, 9:37 PM
mav committed rS323059: Remove unneeded pmap_change_attr() calls..
Remove unneeded pmap_change_attr() calls.
Aug 31 2017, 5:02 PM
mav committed rS323058: Add/polish some defines..
Add/polish some defines.
Aug 31 2017, 4:32 PM
mav committed rS323053: Fix port control for PEX 8749..
Fix port control for PEX 8749.
Aug 31 2017, 1:41 PM
mav accepted D12144: add sysctls for arc shrinking and growing values.

I have no objections. I suppose that code came from Illumos, and there it is usual to modify kernel variables with debugger without sysctls. Adding sysctl would be a good FreeBSD alternative. If somebody wish to shoot his foot -- he should be able to do it.

Aug 31 2017, 11:38 AM
mav added a comment to D12163: [RFC]: decrement needfree by the amount of evicted cache.

On quick look I do see valid point in this patch: needfree in Illumos indeed reduced back as soon as allocation that increased it succeeds, but it doesn't happen in our case. But I am quite busy now to look deeper from all possible sides.

Aug 31 2017, 11:28 AM
mav committed rS323047: Make ntb_set_ctx() always generate fake link event..
Make ntb_set_ctx() always generate fake link event.
Aug 31 2017, 10:59 AM
mav committed rS323046: Make ntb_transport(4) ready receive early link events..
Make ntb_transport(4) ready receive early link events.
Aug 31 2017, 10:53 AM
mav committed rS323045: Fix flags field decoding in ACPI_NFIT_CONTROL_REGION..
Fix flags field decoding in ACPI_NFIT_CONTROL_REGION.
Aug 31 2017, 9:08 AM

Aug 30 2017

mav committed rS323032: Add NTB driver for PLX/Avago/Broadcom PCIe switches..
Add NTB driver for PLX/Avago/Broadcom PCIe switches.
Aug 30 2017, 9:16 PM
mav committed rS323013: MFC r322821: Add missing restart_queue initialization..
MFC r322821: Add missing restart_queue initialization.
Aug 30 2017, 6:28 AM

Aug 29 2017

mav committed rS322996: MFC r322802: Fix off-by-one error when parsing SRAT table..
MFC r322802: Fix off-by-one error when parsing SRAT table.
Aug 29 2017, 7:01 AM

Aug 28 2017

mav committed rS322981: Mask doorbells while processing them..
Mask doorbells while processing them.
Aug 28 2017, 8:00 PM
mav committed rS322980: Fix fake interrupt when set doorbell is unmasked..
Fix fake interrupt when set doorbell is unmasked.
Aug 28 2017, 7:53 PM

Aug 23 2017

mav committed rS322821: Add missing restart_queue initialization..
Add missing restart_queue initialization.
Aug 23 2017, 7:00 PM

Aug 22 2017

mav committed rS322802: Fix off-by-one error when parsing SRAT table..
Fix off-by-one error when parsing SRAT table.
Aug 22 2017, 7:56 PM

Aug 16 2017

mav committed rS322573: MFC r322309, r322316:.
MFC r322309, r322316:
Aug 16 2017, 10:30 AM
mav committed rS322572: MFC r322308: Add new Intel Lewisburg and Union Point chipset PCI IDs..
MFC r322308: Add new Intel Lewisburg and Union Point chipset PCI IDs.
Aug 16 2017, 10:28 AM

Aug 14 2017

mav added inline comments to D12020: LibAliasSetMode error was not detected correctly.
Aug 14 2017, 7:05 PM
mav added inline comments to D12020: LibAliasSetMode error was not detected correctly.
Aug 14 2017, 6:51 PM
mav added inline comments to D12020: LibAliasSetMode error was not detected correctly.
Aug 14 2017, 6:36 PM
mav added a comment to D12020: LibAliasSetMode error was not detected correctly.
In D12020#249291, @mav wrote:

I have another idea. Since this function returns resulted flags, it may be possible to make it return as disabled function which has failed to enable. In such case it would both stay compatible with old code and fixed the compilation errors like that, but slightly complicated check for enable error (if somebody really want it).

Ah, you mean to indicate error by returning with the flag not set in the return flags? Yeah, the error check is a little ugly, but I think that works.

Aug 14 2017, 5:45 PM
mav added a comment to D12020: LibAliasSetMode error was not detected correctly.

Hmm. This procedure is exported in userspace libalias, is it okay to change its signature like that?

Aug 14 2017, 5:07 PM
mav added a comment to D12020: LibAliasSetMode error was not detected correctly.

I don't think this is a correct solution. It looks more like a workaround. Correct IMHO would be to make LibAliasSetMode() return int instead of unsigned it.

Aug 14 2017, 12:35 PM

Aug 12 2017

mav committed rS322423: MFC r322302: Do not loose CCB flags after r320493..
MFC r322302: Do not loose CCB flags after r320493.
Aug 12 2017, 10:22 AM

Aug 10 2017

mav committed rS322351: Add two NFIT fields missed in r321298..
Add two NFIT fields missed in r321298.
Aug 10 2017, 10:59 AM

Aug 9 2017

mav committed rS322316: Missing remanant of 322309..
Missing remanant of 322309.
Aug 9 2017, 1:46 PM
mav committed rS322309: Use "Ibex Peak" codename for "5 Series/3400 Series" chipsets..
Use "Ibex Peak" codename for "5 Series/3400 Series" chipsets.
Aug 9 2017, 12:21 PM
mav committed rS322308: Add new Intel Lewisburg and Union Point chipset PCI IDs..
Add new Intel Lewisburg and Union Point chipset PCI IDs.
Aug 9 2017, 12:03 PM
mav committed rS322302: Do not loose CCB flags after r320493..
Do not loose CCB flags after r320493.
Aug 9 2017, 9:13 AM

Aug 8 2017

mav closed D11904: Fix few issues of LinuxKPI workqueue by committing rS322272: Fix few issues of LinuxKPI workqueue..
Aug 8 2017, 7:36 PM
mav committed rS322272: Fix few issues of LinuxKPI workqueue..
Fix few issues of LinuxKPI workqueue.
Aug 8 2017, 7:36 PM
mav added a comment to D11920: zfsd(8): Close a race condition when onlining a disk paritition.

You haven't included zpool_read_all_labels method into the patch.

Aug 8 2017, 6:05 PM

Aug 7 2017

mav added inline comments to D11904: Fix few issues of LinuxKPI workqueue.
Aug 7 2017, 3:25 PM
mav added inline comments to D11904: Fix few issues of LinuxKPI workqueue.
Aug 7 2017, 2:40 PM
mav committed rS322169: Fix hrtimer_active() in case of cancellation..
Fix hrtimer_active() in case of cancellation.
Aug 7 2017, 2:34 PM
mav closed D11900: Fix hrtimer_active() in case of cancelllation. by committing rS322169: Fix hrtimer_active() in case of cancellation..
Aug 7 2017, 2:34 PM
mav committed rS322138: MFC r321794: Improve FHA locality control for NFS read/write requests..
MFC r321794: Improve FHA locality control for NFS read/write requests.
Aug 7 2017, 7:40 AM
mav committed rS322137: MFC r321794: Improve FHA locality control for NFS read/write requests..
MFC r321794: Improve FHA locality control for NFS read/write requests.
Aug 7 2017, 7:03 AM