21.11.2019 14:39, Kevin Wolf wrote: > Am 21.11.2019 um 11:30 hat Vladimir Sementsov-Ogievskiy geschrieben: >> 21.11.2019 13:28, Vladimir Sementsov-Ogievskiy wrote: >>> 20.11.2019 21:45, Kevin Wolf wrote: >>>> Signed-off-by: Kevin Wolf<kw...@redhat.com> >>> >>> Hmm, allocating 7G will break tests on small disks, for example, >>> on my 2G tmpfs. >>> >>> So, we probably should >>> detect "+qemu-img: Failed to resize underlying file: Could not write zeros >>> for preallocation: No space left on device" >>> errors and skip test. (better skip testcases, but it's not possible for >>> text-comparing tests :( >> >> Or could you just use smaller disks for resize? What is the purpose of such >> a big size? > > I want to exceed a 32 bit byte count for the write_zeroes operation so > that it would break if patch 1 were missing. I guess I could reduce it > to a little over 4 GB, but not less. > > Hm, though that is only for preallocation=off, which shouldn't actually > allocate the space anyway. I could use smaller sizes for falloc and > full, even though I'm not sure if we're still testing these modes with > sizes larger than INT_MAX anywhere. >
Yes, preallocation=off works for me with big disk, so this should work. -- Best regards, Vladimir