Applying the series with these fixes on my side: - patches 2 and 3: commit message fixes, listed in the replies there; - cover letter: "never completes" -> "can hang forever"; - cover letter: "never resubmitted. Any failing allocation leaves them stuck" -> they wait for the next writeback of the page, which may never come; - cover letter: dropped "with EBADF" from the second bug; - cover letter: dropped the sizes, counts and the host name from the test section, and split the first paragraph.
On Mon, 9 Nov 2026 20:06:01 +0100, Denis V. Lunev wrote: > A write that reaches a read-only ploop device never completes: the > writer hangs in D state, and after a buffered write every later sync on > the host hangs behind it until a reboot. Such writes come from a device The write whose allocation fails completes with EIO, and a write to a cluster already in the top delta completes. What hangs are the pios parked on the same BAT page while the failed update was prepared. > - when a cluster allocation or a discard fails, the pios parked on that > BAT page are never resubmitted (patch 2). Any failing allocation > leaves them stuck, not only one on a read-only device. Patch 1 moves The next writeback of that page resubmits them, so after a single -ENOMEM the next successful allocation on the page frees them. They stay for good only when every allocation fails: a read-only device, or no free cluster left. > - ploop test suite, make -j48 test on dev-cl04, green for ploop and The cover letter goes into each commit as the "Patchset description", so the numbers and the host name would end up in the git log. Not for this series: with it the libvzctl backup attach that mounts the partition read-write fails with EIO, and with ploop PR 513 it fails at open. Does libvzctl need a change too, and is there a ticket for it? -- Konstantin Khorenko, Virtuozzo Linux Kernel Team _______________________________________________ Devel mailing list [email protected] https://lists.openvz.org/mailman/listinfo/devel
