On Mon, Aug 19, 2019 at 04:14:11AM +0800, Gao Xiang wrote: > Hi all, > > On Mon, Aug 19, 2019 at 02:16:55AM +0800, Gao Xiang wrote: > > Hi Hch, > > > > On Sun, Aug 18, 2019 at 10:47:02AM -0700, Christoph Hellwig wrote: > > > On Sun, Aug 18, 2019 at 10:29:38AM -0700, Eric Biggers wrote: > > > > Not sure what you're even disagreeing with, as I *do* expect new filesystems to > > > > be held to a high standard, and to be written with the assumption that the > > > > on-disk data may be corrupted or malicious. We just can't expect the bar to be > > > > so high (e.g. no bugs) that it's never been attained by *any* filesystem even > > > > after years/decades of active development. If the developers were careful, the > > > > code generally looks robust, and they are willing to address such bugs as they > > > > are found, realistically that's as good as we can expect to get... > > > > > > Well, the impression I got from Richards quick look and the reply to it is > > > that there is very little attempt to validate the ondisk data structure > > > and there is absolutely no priority to do so. Which is very different > > > from there is a bug or two here and there. > > > > As my second reply to Richard, I didn't fuzz all the on-disk fields for EROFS. > > and as my reply to Richard / Greg, current EROFS is used on the top of dm-verity. > > > > I cannot say how well EROFS will be performed on malformed images (and you can > > also find the bug richard pointed out is a miswritten break->continue by myself). > > > > I posted the upstream EROFS post on July 4, 2019 and a month and a half later, > > no one can tell me (yes, thanks for kind people reply me about their suggestion) > > what we should do next (you can see these emails, I sent many times) to meet > > the minimal upstream requirements and rare people can even dip into my code. > > > > That is all I want to say. I will work on autofuzz these days, and I want to > > know how to meet your requirements on this (you can tell us your standard, > > how well should we do). > > > > OK, you don't reply to my post once, I have no idea how to get your first reply. > > I have made a simple fuzzer to inject messy in inode metadata, > dir data, compressed indexes and super block, > https://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git/commit/?h=experimental-fuzzer > > I am testing with some given dirs and the following script. > Does it look reasonable? > > # !/bin/bash > > mkdir -p mntdir > > for ((i=0; i<1000; ++i)); do > mkfs/mkfs.erofs -F$i testdir_fsl.fuzz.img testdir_fsl > /dev/null 2>&1 mkfs fuzzes the image? Er.... Over in XFS land we have an xfs debugging tool (xfs_db) that knows how to dump (and write!) most every field of every metadata type. This makes it fairly easy to write systematic level 0 fuzzing tests that check how well the filesystem reacts to garbage data (zeroing, randomizing, oneing, adding and subtracting small integers) in a field. (It also knows how to trash entire blocks.) You might want to write such a debugging tool for erofs so that you can take apart crashed images to get a better idea of what went wrong, and to write easy fuzzing tests. --D > umount mntdir > mount -t erofs -o loop testdir_fsl.fuzz.img mntdir > for j in `find mntdir -type f`; do > md5sum $j > /dev/null > done > done > > Thanks, > Gao Xiang > > > > > Thanks, > > Gao Xiang > >
