On 07/14/2016 10:08 PM, Eric Blake wrote: > On 07/08/2016 04:35 PM, Eric Blake wrote: >> Drivers should be able to rely on the block layer honoring the >> max transfer length, rather than needing to return -EINVAL >> (iscsi) or manually fragment things (nbd). This patch adds >> the fragmentation in the block layer, after requests have been >> aligned (fragmenting before alignment would lead to multiple >> unaligned requests, rather than just the head and tail). >> >> The return value was previously nebulous on success (sometimes >> zero, sometimes the length read); since we never have a short >> read, and since fragmenting may store yet another positive >> value in 'ret', change the function to always return the >> incoming 'bytes' value on success. >> >> Signed-off-by: Eric Blake <ebl...@redhat.com> >> >> --- >> v2: Fix uninitialized use of 'ret' for an all-zero read beyond eof > > Uggh. Something I did here and not in v1 is now causing 'make > check-qtest' failures. Please don't merge until I've posted v3.
Looks like there is at least one caller that expects bdrv_aligned_preadv() to return 0 (not positive) on success; I'm not sure which one(s), as it turned into a lot of code to chase, but a simple tweak to guarantee ret = 0 on success solves the failures in 'make check'. v3 coming up soon. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
