Bug 228517 – [4.3/wrong-code] causes ext* filesystems to oops: BUG: unable to handle kernel paging request at <first byte, next page>

• Status: RESOLVED FIXED
• Product: Gentoo Linux

[Description Peter Alfredsen (RETIRED) 2008-06-20 11:55:30 UTC]

Upstream bug: "Incorrectly assumed aligned_operand"
This is a regression from gcc-4.3.0->gcc-4.3.1 introduced by the fix for PR28690: http://gcc.gnu.org/bugzilla/show_bug.cgi?28690
Upstream commit: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=135124

Fedora bug: https://bugzilla.redhat.com/show_bug.cgi?id=451068

[Comment 1 Rafał Mużyło 2008-06-23 19:34:17 UTC]

Well, today they created following upstream commit:
http://gcc.gnu.org/viewcvs?view=rev&revision=137039
How long till it be added to gcc patches ?

[Comment 2 Holger Hoffstätte 2008-06-23 20:01:06 UTC]

FWIW simply compiling the pr36533.c test with -Os will exhibit the bug (segfault). If toolchain/kernel teams are busy I'd be glad to volunteer and provide an ebuild patch.

[Comment 3 Holger Hoffstätte 2008-06-24 08:38:48 UTC]

Created attachment 158189 [details, diff]
Patch for gcc bug #36533

[Comment 4 Holger Hoffstätte 2008-06-24 08:39:34 UTC]

Created attachment 158191 [details, diff]
Patch for ebuild

[Comment 5 Holger Hoffstätte 2008-06-24 08:43:44 UTC]

The patches will fix gcc to correctly compile the test program from http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36533 with -Os.
Other programs still compile & run as well ;)

[Comment 6 Holger Hoffstätte 2008-06-24 08:46:59 UTC]

(In reply to comment #5)
> The patches will fix gcc to correctly compile the test program from
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36533 with -Os.

Forget the above URL - if you need the standalone test use the "upstream commit"  mentioned in comment#1.

[Comment 7 Peter Alfredsen (RETIRED) 2008-07-09 09:16:54 UTC]

Closing, fixed in GCC-4.3.1-r1