Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Control: tag -1 - moreinfo Control: tag -1 fixed-upstream This is claimed to be fixed upstream by: commit e3786b29c54cdae3490b07180a54e2461f42144c Author: Dominique Martinet Date: Thu Aug 8 14:29:38 2024 +0100 9p: Fix DIO read through netfs which is now in Linus's tree and will be part of 6.11-rc4. Ben. -- Ben Hutchings Design a system any fool can use, and only a fool will want to use it. signature.asc Description: This is a digitally signed message part
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Package: linux-image-6.9.7+debian+tj Followup-For: Bug #1072004 X-Debbugs-Cc: [email protected] I've completed 10 successful passes of the autopkgtest reproducer with the proposed patch in the kernel bugzilla from Dominique Martinet on current mainline, so with luck that might squeeze in to 6.10.
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Hi, On 2024-05-30 12:00, Thorsten Leemhuis wrote: > On 29.05.24 23:12, Emanuele Rocca wrote: > > On 2024-05-27 02:31, Thorsten Leemhuis wrote: > >> Would also help a lot to know if this is a 6.8.y only thing, or happens > >> with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. > > > > I could reproduce the issue with 6.9.2 too, will try mainline tomorrow. > > Good to know, thx! Mainline is affected too. > > FTR the procedure is quite simple. [...] > > Don't tell me (I'm just a man-in-the-middle for triaging here and not a > developer for that code), tell David in a reply to > https://lore.kernel.org/all/[email protected]/ – > ideally while answering his question. He needs to know about this. Done.
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
On 29.05.24 23:12, Emanuele Rocca wrote: > On 2024-05-27 02:31, Thorsten Leemhuis wrote: >> Would also help a lot to know if this is a 6.8.y only thing, or happens >> with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. > > I could reproduce the issue with 6.9.2 too, will try mainline tomorrow. Good to know, thx! > FTR the procedure is quite simple. [...] Don't tell me (I'm just a man-in-the-middle for triaging here and not a developer for that code), tell David in a reply to https://lore.kernel.org/all/[email protected]/ – ideally while answering his question. He needs to know about this. Ciao, Thorsten
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Hello Thorsten, On 2024-05-27 02:31, Thorsten Leemhuis wrote: > Would also help a lot to know if this is a 6.8.y only thing, or happens > with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. I could reproduce the issue with 6.9.2 too, will try mainline tomorrow. FTR the procedure is quite simple. To build a sid qemu testbed: autopkgtest-build-qemu unstable /tmp/sid.img Run the autopkgtest: autopkgtest -ddd -B dpdk -- autopkgtest-virt-qemu --debug --show-boot /tmp/sid.img After a while the test hangs with: autopkgtest-virt-qemu: DBG: executing copydown /tmp/alog/tests-tree/ /tmp/autopkgtest.uG6tsJ/build.6QA/src/ [...] autopkgtest-virt-qemu: DBG: +>? Then one can build a new kernel with: make oldconfig scripts/config --disable DEBUG_INFO scripts/config --disable DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT make -j`nproc` bindeb-pkg Boot the QCOW image: qemu-system-x86_64 -cpu host -machine accel=kvm -drive file=sid.img,format=qcow2 -m 4G -smp `nproc` Upgrade the kernel, and run the autopkgtest again as above. Emanuele
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Hi, On 28-05-2024 10:54 a.m., Luca Boccassi wrote: If 6.8 migrates to testing, it will break amd64 debci for unrelated packages for migration tests too. I don't think that's something we want? Paul, wouldn't that qualify as RC? With the kernel team being aware of the issue, I trust the kernel team to balance the options appropriately. I defer to make a call as a Release Team member. Options that I'm aware of: 1) accept the kernel package as is in testing where it will cause some regression in some autopkgtest on ci.d.n infrastructure in testing. We have (only) 56 packages tested with qemu and I could disable qemu based testing on ci.d.n until this bug is resolved (which means that isolation-machine tests will not be run, but also not cause unnecessary failures). 2) update the kernel package with the bisected commit reverted. I understand that the kernel team tries to follow upstream as much as possible to avoid Debian delta's that might be hard to get rid of. 3) update the kernel package with a proposed (but not accepted) patch. 4) wait with updating the kernel package in testing until this issue is solved upstream, causing all kernel fixes to hit testing later. Paul OpenPGP_signature.asc Description: OpenPGP digital signature
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
On Tue, 28 May 2024 at 04:47, Salvatore Bonaccorso wrote: > > Hi Luca, > > On Mon, May 27, 2024 at 09:49:47PM +0100, Luca Boccassi wrote: > > On Mon, 27 May 2024 14:31:37 +0200 Thorsten Leemhuis > > wrote: > > > On 27.05.24 14:22, Luca Boccassi wrote: > > > >> https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/2056461 > > > > > > > > This has been reported upstream 3 weeks ago, but so far it seems no > > > > action has been taken: > > > > > > > > https://lore.kernel.org/all/Zj0ErxVBE3DYT2Ea@gpd/ > > > > > > Hmmm, that thread is strange, why are David's replies not where they > > are > > > supposed to be? Whatever. The last thing from just a few days ago > > seems > > > to be a inquiry from David to Andrea that was not yet answered > > afaics: > > > https://lore.kernel.org/all/[email protected]/ > > > > > > Would also help a lot to know if this is a 6.8.y only thing, or > > happens > > > with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. > > > > That's not available anywhere (in a usable form) so it's a question for > > kernel maintainers. > > > > Salvatore, what's the moreinfo for? Were you not able to reproduce it? > > Just try to run any isolation-machine autopkgtest (eg: src:dpdk) in a > > qem image (eg: built with autopkgtest-build-qemu) and compare unstable > > vs testing. > > The aim for the moreinfo tag is to gather more information on which if > we know it's fixed with 6.9.y or mainline to have an idea why the > above thread might have stalled upstream. > > I think I would ideally have first one 6.8.y version going to testing > and Diederik did work on rebasing already for 6.9.y, so we might move > then to that one with 6.8.y going EOL. If 6.8 migrates to testing, it will break amd64 debci for unrelated packages for migration tests too. I don't think that's something we want? Paul, wouldn't that qualify as RC?
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Hi Luca, On Mon, May 27, 2024 at 09:49:47PM +0100, Luca Boccassi wrote: > On Mon, 27 May 2024 14:31:37 +0200 Thorsten Leemhuis > wrote: > > On 27.05.24 14:22, Luca Boccassi wrote: > > >> https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/2056461 > > > > > > This has been reported upstream 3 weeks ago, but so far it seems no > > > action has been taken: > > > > > > https://lore.kernel.org/all/Zj0ErxVBE3DYT2Ea@gpd/ > > > > Hmmm, that thread is strange, why are David's replies not where they > are > > supposed to be? Whatever. The last thing from just a few days ago > seems > > to be a inquiry from David to Andrea that was not yet answered > afaics: > > https://lore.kernel.org/all/[email protected]/ > > > > Would also help a lot to know if this is a 6.8.y only thing, or > happens > > with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. > > That's not available anywhere (in a usable form) so it's a question for > kernel maintainers. > > Salvatore, what's the moreinfo for? Were you not able to reproduce it? > Just try to run any isolation-machine autopkgtest (eg: src:dpdk) in a > qem image (eg: built with autopkgtest-build-qemu) and compare unstable > vs testing. The aim for the moreinfo tag is to gather more information on which if we know it's fixed with 6.9.y or mainline to have an idea why the above thread might have stalled upstream. I think I would ideally have first one 6.8.y version going to testing and Diederik did work on rebasing already for 6.9.y, so we might move then to that one with 6.8.y going EOL. Regards, Salvatore
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
On Mon, 27 May 2024 14:31:37 +0200 Thorsten Leemhuis wrote: > On 27.05.24 14:22, Luca Boccassi wrote: > >> https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/2056461 > > > > This has been reported upstream 3 weeks ago, but so far it seems no > > action has been taken: > > > > https://lore.kernel.org/all/Zj0ErxVBE3DYT2Ea@gpd/ > > Hmmm, that thread is strange, why are David's replies not where they are > supposed to be? Whatever. The last thing from just a few days ago seems > to be a inquiry from David to Andrea that was not yet answered afaics: > https://lore.kernel.org/all/[email protected]/ > > Would also help a lot to know if this is a 6.8.y only thing, or happens > with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. That's not available anywhere (in a usable form) so it's a question for kernel maintainers. Salvatore, what's the moreinfo for? Were you not able to reproduce it? Just try to run any isolation-machine autopkgtest (eg: src:dpdk) in a qem image (eg: built with autopkgtest-build-qemu) and compare unstable vs testing. -- Kind regards, Luca Boccassi signature.asc Description: This is a digitally signed message part
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
On 27.05.24 14:22, Luca Boccassi wrote: >> https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/2056461 > > This has been reported upstream 3 weeks ago, but so far it seems no > action has been taken: > > https://lore.kernel.org/all/Zj0ErxVBE3DYT2Ea@gpd/ Hmmm, that thread is strange, why are David's replies not where they are supposed to be? Whatever. The last thing from just a few days ago seems to be a inquiry from David to Andrea that was not yet answered afaics: https://lore.kernel.org/all/[email protected]/ Would also help a lot to know if this is a 6.8.y only thing, or happens with 6.9 and mainline as well, as 6.8.y will likely be EOLed soon. Ciao, Thorsten
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
On Mon, 27 May 2024 13:02:12 +0100 Luca Boccassi wrote: > Source: linux > Version: 6.8.9-1 > Severity: grave > Justification: breaks autopkgtest jobs running in qemu, breaking amd64 debci > X-Debbugs-CC: [email protected], [email protected] > > Hi, > > Kernel 6.8 includes this change: > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=80105ed2fd2715fb09a8fdb0655a8bdc86c120db > > This has been bisected by Canonical and found to be breaking > qemu+autopkgtest. Tests just hang while waiting for input. > > This is reliably reproducible, and has already started affecting debci > amd64 jobs in unstable, so filing with high severity to avoid > migration, that would break migration debci autopkgtest jobs. Example: > > https://ci.debian.net/packages/s/systemd/unstable/amd64/47041978/ > > The launchpad ticket has more details: > > https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/2056461 This has been reported upstream 3 weeks ago, but so far it seems no action has been taken: https://lore.kernel.org/all/Zj0ErxVBE3DYT2Ea@gpd/ Maybe Thorsten could be able to help? TIA! -- Kind regards, Luca Boccassi signature.asc Description: This is a digitally signed message part
Bug#1072004: linux: regression in the 9p protocol in 6.8 breaks autopkgtest qemu jobs (affecting debci)
Source: linux Version: 6.8.9-1 Severity: grave Justification: breaks autopkgtest jobs running in qemu, breaking amd64 debci X-Debbugs-CC: [email protected], [email protected] Hi, Kernel 6.8 includes this change: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=80105ed2fd2715fb09a8fdb0655a8bdc86c120db This has been bisected by Canonical and found to be breaking qemu+autopkgtest. Tests just hang while waiting for input. This is reliably reproducible, and has already started affecting debci amd64 jobs in unstable, so filing with high severity to avoid migration, that would break migration debci autopkgtest jobs. Example: https://ci.debian.net/packages/s/systemd/unstable/amd64/47041978/ The launchpad ticket has more details: https://bugs.launchpad.net/ubuntu/+source/autopkgtest/+bug/2056461 -- Kind regards, Luca Boccassi signature.asc Description: This is a digitally signed message part
