From: Claus-Justus Heine <ch@dot-heine.de>
To: linux-kernel@vger.kernel.org
Subject: Re: 2.6.0-test4: bug WRT lazy FPU switching and CONFIG_PREEMPT
Date: Mon, 01 Sep 2003 15:15:50 +0200 [thread overview]
Message-ID: <3F534686.1090106@dot-heine.de> (raw)
In-Reply-To: <qCCS.70v.9@gated-at.bofh.it>
Claus-Justus Heine wrote:
[snip]
> diff -u --recursive --new-file linux-2.6.0-test4/include/asm-i386/i387.h linux-2.6.0-test4-mine/include/asm-i386/i387.h
[snip]
> --- linux-2.6.0-test4/include/asm-i386/i387.h 2003-07-27 19:06:54.000000000 +0200
> +++ linux-2.6.0-test4-mine/include/asm-i386/i387.h 2003-08-31 15:39:04.000000000 +0200
> @@ -30,7 +30,7 @@
> static inline void __save_init_fpu( struct task_struct *tsk )
> {
> if ( cpu_has_fxsr ) {
> - asm volatile( "fxsave %0 ; fnclex"
> + asm volatile( "fxsave %0 ; fclex"
I'm sorry, changing "fnclex" to "fclex" was a stupid idea, can lead to deadlocks. So that
patch should have looked like this:
#########################################################################################################################
diff -u --recursive --new-file linux-2.6.0-test4/arch/i386/kernel/traps.c linux-2.6.0-test4-mine/arch/i386/kernel/traps.c
--- linux-2.6.0-test4/arch/i386/kernel/traps.c 2003-08-31 15:22:42.000000000 +0200
+++ linux-2.6.0-test4-mine/arch/i386/kernel/traps.c 2003-08-31 15:35:49.000000000 +0200
@@ -605,7 +605,10 @@
* Save the info for the exception handler and clear the error.
*/
task = current;
- save_init_fpu(task);
+ /* don't trigger an unnecessary math_state_restore() */
+ if (task->thread_info->status & TS_USEDFPU) {
+ save_init_fpu(task);
+ }
task->thread.trap_no = 16;
task->thread.error_code = 0;
info.si_signo = SIGFPE;
@@ -667,7 +670,10 @@
* Save the info for the exception handler and clear the error.
*/
task = current;
- save_init_fpu(task);
+ /* don't trigger an unnecessary math_state_restore() */
+ if (task->thread_info->status & TS_USEDFPU) {
+ save_init_fpu(task);
+ }
task->thread.trap_no = 19;
task->thread.error_code = 0;
info.si_signo = SIGFPE;
diff -u --recursive --new-file linux-2.6.0-test4/include/asm-i386/i387.h linux-2.6.0-test4-mine/include/asm-i386/i387.h
--- linux-2.6.0-test4/include/asm-i386/i387.h 2003-07-27 19:06:54.000000000 +0200
+++ linux-2.6.0-test4-mine/include/asm-i386/i387.h 2003-08-31 15:39:04.000000000 +0200
@@ -41,8 +41,10 @@
static inline void save_init_fpu( struct task_struct *tsk )
{
+ preempt_disable();
__save_init_fpu(tsk);
stts();
+ preempt_enable_no_resched();
}
@@ -53,11 +55,13 @@
#define clear_fpu( tsk ) \
do { \
+ preempt_disable(); \
if ((tsk)->thread_info->status & TS_USEDFPU) { \
asm volatile("fwait"); \
(tsk)->thread_info->status &= ~TS_USEDFPU; \
stts(); \
} \
+ preempt_enable_no_resched(); \
} while (0)
/*
##########################################################################################################################################
next parent reply other threads:[~2003-09-01 13:15 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <qCCS.70v.9@gated-at.bofh.it>
2003-09-01 13:15 ` Claus-Justus Heine [this message]
2003-08-31 15:42 2.6.0-test4: bug WRT lazy FPU switching and CONFIG_PREEMPT Claus-Justus Heine
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3F534686.1090106@dot-heine.de \
--to=ch@dot-heine.de \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox