On Tue, Jan 14, 2020 at 4:53 PM Eric Biggers <ebiggers@xxxxxxxxxx> wrote: > > On Tue, Jan 07, 2020 at 12:28:25PM +0000, Joe Thornber wrote: > > On Tue, Jan 07, 2020 at 10:46:27AM +0000, Joe Thornber wrote: > > > I'll get a patch to you later today. > > > > Eric, > > > > Patch below. I've run it through a bunch of tests in the dm test suite. But > > obviously I have never hit your issue. Will do more testing today. > > > > - Joe > > > > > > > > Author: Joe Thornber <ejt@xxxxxxxxxx> > > Date: Tue Jan 7 11:58:42 2020 +0000 > > > > [dm-thin, dm-cache] Fix bug in space-maps. > > > > The space-maps track the reference counts for disk blocks. There are variants > > for tracking metadata blocks, and data blocks. > > > > We implement transactionality by never touching blocks from the previous > > transaction, so we can rollback in the event of a crash. > > > > When allocating a new block we need to ensure the block is free (has reference > > count of 0) in both the current and previous transaction. Prior to this patch we > > were doing this by searching for a free block in the previous transaction, and > > relying on a 'begin' counter to track where the last allocation in the current > > transaction was. This 'begin' field was not being updated in all code paths (eg, > > increment of a data block reference count due to breaking sharing of a neighbour > > block in the same btree leaf). > > > > This patch keeps the 'begin' field, but now it's just a hint to speed up the search. > > Instead we search the current transaction for a free block, and then double check > > it's free in the old transaction. Much simpler. > > > > I happened to notice this patch is on the linux-dm/for-next branch > (https://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm.git/commit/?h=for-next&id=2137c0dcc04b24efb4c38d4b46b7194575718dd5) > and it has: > > Reported-by: Eric Biggers <ebiggers@xxxxxxxxxx> > > This is wrong, I didn't report this. I think you meant to put: > > Reported-by: Eric Wheeler <dm-devel@xxxxxxxxxxxxxxxxxx> > > - Eric (the other one) Fixed it up, not sure how that happened, sorry about that!
