On Mon, Jul 13, 2020 at 11:16:10AM +0200, Carlos Maiolino wrote: > Hi Dave, Christoph. > > > > > - ip = kmem_zone_alloc(xfs_inode_zone, 0); > > > > + > > > > + if (current->flags & PF_MEMALLOC_NOFS) > > > > + gfp_mask |= __GFP_NOFAIL; > > > > > > I'm a little worried about this change in beavior here. Can we > > > just keep the unconditional __GFP_NOFAIL and if we really care do the > > > change separately after the series? At that point it should probably > > > use the re-added PF_FSTRANS flag as well. > > > Checking PF_FSTRANS was what I suggested should be done here, not > > PF_MEMALLOC_NOFS... > > > No problem in splitting this change into 2 patches, 1 by unconditionally use > __GFP_NOFAIL, and another changing the behavior to use NOFAIL only inside a > transaction. > > Regarding the PF_FSTRANS flag, I opted by PF_MEMALLOC_NOFS after reading the > commit which removed PF_FSTRANS initially (didn't mean to ignore your suggestion > Dave, my apologies if I sounded like that), but I actually didn't find any commit > re-adding PF_FSTRANS back. I searched most trees but couldn't find any commit > re-adding it back, could you guys please point me out where is the commit adding > it back? I suspect Dave is referring to: "xfs: reintroduce PF_FSTRANS for transaction reservation recursion protection" by Yang Shao. AFAICT it hasn't cleared akpm yet, so it's not in his quiltpile, and as he doesn't use git there won't be a commit until it ends up in mainline... --D > > > > Cheers, > > > > Dave. > > -- > > Dave Chinner > > david@xxxxxxxxxxxxx > > > > -- > Carlos >
