* [parisc-linux] 2.5 fix for user level page fault problems
@ 2002-12-18 16:20 James Bottomley
2002-12-18 16:34 ` Randolph Chung
0 siblings, 1 reply; 2+ messages in thread
From: James Bottomley @ 2002-12-18 16:20 UTC (permalink / raw)
To: parisc-linux; +Cc: James.Bottomley
[-- Attachment #1: Type: text/plain, Size: 534 bytes --]
It seems that going from 2.4 to 2.5 there was an over zealous removal of the
check signals return path. The problem is that if a user application takes a
page fault, that can result in a signal being posted (specifically SEGV for
illegal memory access). If we never check the signals, the instruction is
retried and re-faults ad infinitum (well actually, it seems to terminate with
an unaligned instruction trap in this case after a few hundred faults).
The attached path makes signal posting on page fault work again
James
[-- Attachment #2: tmp.diff --]
[-- Type: text/plain , Size: 422 bytes --]
===== entry.S 1.8 vs edited =====
--- 1.8/arch/parisc/kernel/entry.S Sun Nov 24 17:36:53 2002
+++ edited/entry.S Mon Dec 16 22:09:59 2002
@@ -944,11 +944,11 @@
ldo -16(%r30),%r29 /* Reference param save area */
#endif
- ldil L%intr_restore, %r2
+ ldil L%intr_check_sig, %r2
copy %r25, %r16 /* save pt_regs */
b handle_interruption
- ldo R%intr_restore(%r2), %r2
+ ldo R%intr_check_sig(%r2), %r2
/*
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [parisc-linux] 2.5 fix for user level page fault problems
2002-12-18 16:20 [parisc-linux] 2.5 fix for user level page fault problems James Bottomley
@ 2002-12-18 16:34 ` Randolph Chung
0 siblings, 0 replies; 2+ messages in thread
From: Randolph Chung @ 2002-12-18 16:34 UTC (permalink / raw)
To: James Bottomley; +Cc: parisc-linux
In reference to a message from James Bottomley, dated Dec 18:
> It seems that going from 2.4 to 2.5 there was an over zealous removal of the
> check signals return path. The problem is that if a user application takes a
> page fault, that can result in a signal being posted (specifically SEGV for
> illegal memory access). If we never check the signals, the instruction is
> retried and re-faults ad infinitum (well actually, it seems to terminate with
> an unaligned instruction trap in this case after a few hundred faults).
thanks James. I've commited this to our 2.5 tree. We had fixed this in
2.4 a couple of weeks ago but I forgot to commit this to 2.5 too.
randolph
--
Randolph Chung
Debian GNU/Linux Developer, hppa/ia64 ports
http://www.tausq.org/
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2002-12-18 16:31 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-12-18 16:20 [parisc-linux] 2.5 fix for user level page fault problems James Bottomley
2002-12-18 16:34 ` Randolph Chung
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.