On Thu, Apr 18, 2019 at 01:19:10PM +0200, Miklos Szeredi wrote: > On Wed, Apr 17, 2019 at 10:04 PM Liu Bo <bo.liu@xxxxxxxxxxxxxxxxx> wrote: > > > > From: Xiaoguang Wang <xiaoguang.wang@xxxxxxxxxxxxxxxxx> > > > > If page straddles i_size and we write the whole page, the fuse > > user-space filesystem may extend file size, it will confuse users. > > > > Before this patch: > > xfs_io -t -f \ > > -c "truncate 5120" \ > > -c "pwrite -S 0x58 0 5120" \ > > -c "mmap -rw 0 5120" \ > > -c "mwrite -S 0x59 2048 3072" \ > > -c "close" \ > > testfile > > testfile's length will be 8192 bytes, with this patch, testfile's > > length will be 5120 bytes. > > Cannot reproduce this. The file size should be cropped in > fuse_flush_writepages() and the above test case produces the expected > result, so I'm not sure how were you able to get 8192 byte file. Hmm, this was found by fstests while testing against virtiofs 0.1 daemon code, which doesn't handle file size from fuse_in_head. but now with 0.2, /me couldn't reproduce it, either. Please ignore the patch as well. thanks, -liubo
