| Message ID | 20160907083107.GC5113@noname.redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hello Eric, Kevin, unfortunately it was not possible for me to get the git tree at this time. I have insert the changes by hand, which was into this mail below in block/iscsi.c in the distribution from qemu.org for 2.7 (was this o.k.)? After installing and start with virsh create, the following problem occur: internal error: process exited while connecting to monitor: qemu-system-x86_64: block.c:1022: bdrv_open_common: Assertion `is_power_of_2(bs->bl.request_alignment)' failed. We will send you the backtrace later on. Best regards Holger Am 07.09.2016 um 10:31 schrieb Kevin Wolf: > Am 06.09.2016 um 21:04 hat Eric Blake geschrieben: >> When qemu uses iscsi devices in sg mode, iscsilun->block_size >> is left at 0. Prior to commits cf081fca and similar, when >> block limits were tracked in sectors, this did not matter: >> various block limits were just left at 0. But when we started >> scaling by block size, this caused SIGFPE. >> >> One possible solution for SG mode is to just blindly skip ALL >> of iscsi_refresh_limits(), since we already short circuit so >> many other things in sg mode. But this patch takes a slightly >> more conservative approach, and merely guarantees that scaling >> will succeed (for SG devices, the scaling is done to the block >> layer default of 512 bytes, since we don't know of any iscsi >> devices with a smaller block size), while still using multiples >> of the original size. Resulting limits may still be zero in SG >> mode (that is, we only fix block_size used as a denominator, not >> all uses). >> >> Reported-by: Holger Schranz <holger@fam-schranz.de> >> Signed-off-by: Eric Blake <eblake@redhat.com> >> CC: qemu-stable@nongnu.org >> --- >> >> I would really appreciate Holger testing this patch. We could >> also go with the much shorter patch that just does >> if (bs->sg) { return; } >> at the top of iscsi_refresh_limits(), but I'm not sure if that >> would break anything else in the block layer (we had several, >> but not all, limits that were provably left alone at 0 for >> sg mode). > Hm, originally I thought that we could even just return for bs->sg in > bdrv_refresh_limits() like below because for SG devices the limits > should never be used, but with scsi-block they might actually be. > > Paolo, what do you think? > > Anyway, let's go with the above patch first, it's a conservative one > that qemu-stable and possibly downstream can safely backport. If we're > going to add anything else, that would be for 2.8 only. > > Kevin > > diff --git a/block/io.c b/block/io.c > index fdf7080..144ff65 100644 > --- a/block/io.c > +++ b/block/io.c > @@ -84,7 +84,7 @@ void bdrv_refresh_limits(BlockDriverState *bs, Error **errp) > > memset(&bs->bl, 0, sizeof(bs->bl)); > > - if (!drv) { > + if (!drv || bs->sg) { > return; > } > --- Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft. https://www.avast.com/antivirus
On 09/07/2016 06:48 AM, Holger Schranz wrote: > Hello Eric, Kevin, > > unfortunately it was not possible for me to get the git tree at this > time. I have insert the changes by hand, which was into this mail below > in block/iscsi.c in the distribution from qemu.org for 2.7 (was this o.k.)? Probably good enough, as it at least got you further (doesn't feel quite as clean as testing git directly, but since the patch should be backportable as-is to the 2.7 code, it matches what downstream distros will be doing). > > After installing and start with virsh create, the following problem occur: > > internal error: process exited while connecting to monitor: > qemu-system-x86_64: block.c:1022: > bdrv_open_common: Assertion `is_power_of_2(bs->bl.request_alignment)' > failed. > > We will send you the backtrace later on. Ah, I already know where the backtrace would point to. The patch is setting bl.request_alignment to 0, which is not a power of two, so the assertion in block.c should also exempt SG devices from reporting a preferred request_alignment. v2 coming up.
diff --git a/block/io.c b/block/io.c index fdf7080..144ff65 100644 --- a/block/io.c +++ b/block/io.c @@ -84,7 +84,7 @@ void bdrv_refresh_limits(BlockDriverState *bs, Error **errp) memset(&bs->bl, 0, sizeof(bs->bl)); - if (!drv) { + if (!drv || bs->sg) { return; }