On 2020/5/6 14:55, Chao Yu wrote: > On 2020/5/6 9:24, Eric Biggers wrote: >> On Wed, May 06, 2020 at 08:14:07AM +0800, Gao Xiang wrote: >>>> >>>> Actually, I think this is wrong because the fsync can be done via a file >>>> descriptor that was opened to a now-deleted link to the file. >>> >>> I'm still confused about this... >>> >>> I don't know what's wrong with this version from my limited knowledge? >>> inode itself is locked when fsyncing, so >>> >>> if the fsync inode->i_nlink == 1, this inode has only one hard link >>> (not deleted yet) and should belong to a single directory; and >>> >>> the only one parent directory would not go away (not deleted as well) >>> since there are some dirents in it (not empty). >>> >>> Could kindly explain more so I would learn more about this scenario? >>> Thanks a lot! >> >> i_nlink == 1 just means that there is one non-deleted link. There can be >> links >> that have since been deleted, and file descriptors can still be open to them. >> >>> >>>> >>>> We need to find the dentry whose parent directory is still exists, i.e. the >>>> parent directory that is counting towards 'inode->i_nlink == 1'. >>> >>> directory counting towards 'inode->i_nlink == 1', what's happening? >> >> The non-deleted link is the one counted in i_nlink. >> >>> >>>> >>>> I think d_find_alias() is what we're looking for. >>> >>> It may be simply dentry->d_parent (stable/positive as you said before, and >>> it's >>> not empty). why need to d_find_alias()? >> >> Because we need to get the dentry that hasn't been deleted yet, which isn't >> necessarily the one associated with the file descriptor being fsync()'ed. >> >>> And what is the original problem? I could not get some clue from the >>> original >>> patch description (I only saw some extra igrab/iput because of some unknown >>> reasons), it there some backtrace related to the problem? >> >> The problem is that i_pino gets set incorrectly. I just noticed this while >> reviewing the code. It's not hard to reproduce, e.g.: >> >> #include <unistd.h> >> #include <fcntl.h> >> #include <sys/stat.h> >> >> int main() >> { >> int fd; >> >> mkdir("dir1", 0700); >> mkdir("dir2", 0700); >> mknod("dir1/file", S_IFREG|0600, 0); >> link("dir1/file", "dir2/file"); >> fd = open("dir2/file", O_WRONLY); >> unlink("dir2/file"); >> write(fd, "X", 1); >> fsync(fd); >> } >> >> Then: >> >> sync >> echo N | dump.f2fs -i $(stat -c %i dir1/file) /dev/vdb | grep 'i_pino' >> echo "dir1 (correct): $(stat -c %i dir1)" >> echo "dir2 (wrong): $(stat -c %i dir2)" >> >> i_pino will point to dir2 rather than dir1 as expected. > > Could you add above testcase into commit message of your patch? it will > be easier to understand the issue we solved with it. > > In addition, how about adding this testcase in fstest as a generic one?
Oops, it's not a generic one... please ignore this. > >> >> - Eric >> >> >> _______________________________________________ >> Linux-f2fs-devel mailing list >> linux-f2fs-de...@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel >> . >> > > > _______________________________________________ > Linux-f2fs-devel mailing list > linux-f2fs-de...@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel > . >