[Toybox] [PATCH] lsattr/chattr: add missing flags.

[Rob Landley]

On 1/21/20 3:15 PM, enh wrote:
> On Mon, Jan 20, 2020 at 9:23 AM Rob Landley <rob at landley.net> wrote:
>>
>> On 1/19/20 6:58 PM, enh via Toybox wrote:
>>> Also update help to include flags that were already added. Remove
>>> useless duplicative comments. Swich to FLAG macros.
>>
>> Applied, but I have a TODO item to see if I should migrate all this to FS_XFLAG
>> out of linux/fs.h, ala FS_IOC_FSGETXATTR and friends.
>>
>> (My understanding is the current stuff is ext2/3/4-specific, and the other
>> things apply to more filesystems? I think? It's still a todo for a reason. Green
>> elf needs tests cases, badly.)
> 
> a grep of the kernel gets more matches for FS_IOC_SETFLAGS than for
> for FS_IOC_FSSETXATTR, and the former is available all the way back to
> 2.6 whereas the latter is new in 4.5. so i think the "applicable to
> more filesystems" thing is still mostly theoretical and the 7 year
> rule means we need FS_IOC_SETFLAGS for the foreseeable future anyway?
> 
> i actually assumed your TODO was about the ad hoc argument parsing!
> 
> i do need FS_IOC_FSSETXATTR though to add `-p projid`. i'll send you
> the patch for that when someone with a suitable fs tells me it
> actually works. (i've asked them if there's anything we can do about
> regression testing chattr too, though i'm assuming this needs us to
> mount a sample fs we're prepared to modify.)

A nice person emailed me an xattr test fs and I made a todo item to add xattr
support to tar. It's not hard, just haven't done it yet: it _seems_ to be just
caring about the contents of the 'x' records looking for, but I can't get binary
equivalence unless I figure out why "sub3" is becoming "./PaxHeaders.15625/sub3"
in the x record it creates? It's not an inode number...

Oh: https://www.gnu.org/software/tar/manual/html_chapter/tar_8.html

It's the process ID of the tar process. So they GRATUITOUSLY prevent binary
reproducibility whenever xattrs are involved, EVEN FROM TWO CONSECUTIVE RUNS OF
THE SAME TOO ON THE SAME DATA YOU....

Ahem. Right. Moving on. (Add another --stop-doing-that in the test command line.)

In theory tmpfs mounts support xattrs so I can extract a tarball into a tmpfs to
get reproducible behavior. Still requires root to test, but that should be
reproducible.

In practice I tried to use /dev/shm on devuan and it gave me "inappropriate
ioctl for device while reading flags" so as far as I can tell they haven't even
compiled xattr support into tmpfs in the distro kernel...

Rob