Date: Tue, 05 Jan 2016 23:53:26 +0000 From: bugzilla-noreply@freebsd.org To: freebsd-fs@FreeBSD.org Subject: [Bug 205938] [ext2fs][patch][panic] EXT4: reading mmaped file causes panic because struct buf leaks Message-ID: <bug-205938-3630@https.bugs.freebsd.org/bugzilla/>
next in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D205938 Bug ID: 205938 Summary: [ext2fs][patch][panic] EXT4: reading mmaped file causes panic because struct buf leaks Product: Base System Version: 11.0-CURRENT Hardware: Any OS: Any Status: New Keywords: crash, patch Severity: Affects Many People Priority: --- Component: kern Assignee: freebsd-bugs@FreeBSD.org Reporter: damjan.jov@gmail.com CC: freebsd-fs@FreeBSD.org Created attachment 165127 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=3D165127&action= =3Dedit Fix a kernel panic when reading mmaped files from EXT4 Calling mmap() on any sizeable file on an EXT4 filesystem, and then attempt= ing to read that memory (can be easily tested using the "cmp file file" tool), causes a reproducible kernel panic: userret: returning with the following locks held: exclusive lockmgr bufwait (bufwait) r =3D 0 (0xfffffe001d90c220) locked @ /usr/src/sys/kern/vfs_bio.c:1454 panic: witness_warn cpuid =3D 0 KDB: stack backtrace: db_trace_self_wrapper() at db_trace-self_wrapper+0x2b/frame 0xfffffe002b7e6= 7f0 vpanic() at vpanic+0x182/frame 0xfffffe002b7e6870 kassert_panic() at kassert_panic+0x126/frame 0xfffffe002b7e68e0 witness_warn() at witness_warn+0x3c6/frame 0xfffffe002b7e69b0 userret() at userret+0x98/frame 0xfffffe002b7e69e0 trap() at trap+0x3f4/frame 0xfffffe002b7e6bf0 calltrap() at calltrap+0x8/frame 0xfffffe002b7e6bf0 --- trap 0xc, rip =3D 0x4019c0, rsp =3D 0x7fffffffe940, rbp =3D 0x7ffffffff= eea30 --- KDB: enter: panic [ thread pid 909 tid 100082 ] Stopped at kdb_enter+0x3b: movq $0,kdb_why The problem comes from ext4_bmapext() in sys/fs/ext2fs/ext2_bmap.c never calling brelse(), meaning the "struct buf" returned in path.ep_bp from ext4_ext_find_extent() is never released/unlocked, something userret() catc= hes later and panics from. The attached patch always calls brelse(path.ep_bp), fixing reading EXT4 fil= es using mmap(). This affects all versions of FreeBSD. --=20 You are receiving this mail because: You are on the CC list for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-205938-3630>