Thu, 3 May 2018 06:45:22 +0200 Sebastien Marie <[email protected]>
> On Thu, May 03, 2018 at 12:27:05AM +0300, [email protected] wrote:
> > >
> > > So if i understand, in that case 37 is the fd for places.sqlite-wal
> > > opened earlier in the ktrace ?
> > >
Hi Landry, Sebastien,
Yes, wiped that temp file, but generated a fresh one, different fd num:
50653 firefox CALL clock_gettime(CLOCK_MONOTONIC,0x7f7ffffe1d58)
50653 firefox STRU struct timespec { 44122.094371935 }
50653 firefox RET clock_gettime 0
50653 firefox CALL getpid()
50653 firefox RET getpid 50653/0xc5dd
50653 firefox CALL stat(0x7f7ffffe1b30,0x7f7ffffe18a0)
50653 firefox NAMI "/home/user/.mozilla/firefox/test/places.sqlite"
50653 firefox STRU struct stat { dev=1051, ino=181924, mode=-rw-r--r-- ,
nlink=1, uid=1000<"user">, gid=1000<"user">, rdev=742816, atime=1524794989<"Apr
27 05:09:49 2018">.487312718, mtime=1525294393<"May 2 23:53:13
2018">.783442474, ctime=1525294393<"May 2 23:53:13 2018">.783442474,
size=5242880, blocks=10272, blksize=16384, flags=0x0, gen=0x0 }
50653 firefox RET stat 0
50653 firefox CALL
open(0xb1f8870275d,0x10202<O_RDWR|O_CREAT|O_CLOEXEC>,0644<S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH>)
50653 firefox NAMI "/home/user/.mozilla/firefox/test/places.sqlite-wal"
50653 firefox RET open 40/0x28
50653 firefox CALL fstat(40,0x7f7ffffe1780)
50653 firefox STRU struct stat { dev=1051, ino=182010, mode=-rw-r----- ,
nlink=1, uid=1000<"user">, gid=1000<"user">, rdev=0, atime=1525294428<"May 2
23:53:48 2018">.294304583, mtime=1525294428<"May 2 23:53:48 2018">.294304583,
ctime=1525294428<"May 2 23:53:48 2018">.294304583, size=0, blocks=0,
blksize=16384, flags=0x0, gen=0x0 }
50653 firefox RET fstat 0
50653 firefox CALL fchmod(40,0644<S_IRUSR|S_IWUSR|S_IRGRP|S_IROTH>)
50653 firefox PLDG fchmod, "fattr", errno 1 Operation not permitted
50653 firefox PSIG SIGABRT SIG_DFL
50653 firefox NAMI "firefox.core"
50653 firefox STRU struct pollfd [2] { fd=16, events=0x3<POLLIN|POLLPRI>,
revents=0<> } { fd=37, events=0x3<POLLIN|POLLPRI>, revents=0<> }
50653 firefox STRU struct pollfd { fd=4, events=0x1<POLLIN>, revents=0<> }
Same trace with the stat structs obviously the details are significant.
Please advise if you'd need further traces and how to obtain more info.
> I think I already seen this behaviour in fossil:
> - openbsd ports@ list:
> https://marc.info/?l=openbsd-ports&m=152196264519554&w=2
> - fossil-scm list:
> https://www.mail-archive.com/[email protected]/msg27158.html
>
> The fchmod(2) code behave to libsqlite.
>
> In the fossil case (but I assume similarity here), libsqlite sets the
> mode on "foo-journal" to be the same than "foo". This code path could be
> triggered when the initial run that create "foo" had difference umask
> setting than the current run that create "foo-journal".
>
> Anton, what are the mode of (it was part of stripped stat information :-])
> - /home/user/.mozilla/firefox/test/places.sqlite
> - /home/user/.mozilla/firefox/test/places.sqlite-wal (once created)
Included above, indeed the places.sqlite-wal file has a different mode.
Does firefox use own routines for fattr operations, or the system ones?
And probably the same question for both (lib)sqlite and fossil applies.
> and what is your umask value in this shell ?
The umask is set in the .profile (ksh) as umask 027 (changed from 022).
Is it the right time to precise the pledge path to the profile dir only
and probably another code path would be the save files routines (ouch)?
I suggest web clients should not have access to paths outside $workdir.
Kind regards,
Anton Lazarov
