Re: [BUG] 2.6.17-rc3 broke FP exceptions on x86

[Chuck Ebbert]

In-Reply-To: <17490.4240.578291.222262@xxxxxxxxxxxxxxx>

On Fri, 28 Apr 2006 14:54:40 +0200, Mikael Pettersson wrote:

> Running an FP exception using user-space application
> (the runtime system for the Erlang programming language
> in my case) on an Athlon64 with a 32-bit 2.6.17-rc3 kernel
> quickly results in a complete system hang: mouse is dead,
> keyboard is dead, the network doesn't reply to pings.
> Had to reboot via the power switch to get the machine back.
> 
> This happended twice in a row. With 2.6.17-rc2 things
> work fine like they always have before.

This should fix it... please test.

Signed-off-by: Chuck Ebbert <76306.1226@xxxxxxxxxxxxxx>

--- 2.6.17-rc3-d4.orig/include/asm-i386/i387.h
+++ 2.6.17-rc3-d4/include/asm-i386/i387.h
@@ -58,13 +58,13 @@ static inline void __save_init_fpu( stru
 	alternative_input(
 		"fnsave %[fx] ;fwait;" GENERIC_NOP8 GENERIC_NOP4,
 		"fxsave %[fx]\n"
-		"bt $7,%[fsw] ; jc 1f ; fnclex\n1:",
+		"bt $7,%[fsw] ; jnc 1f ; fnclex\n1:",
 		X86_FEATURE_FXSR,
 		[fx] "m" (tsk->thread.i387.fxsave),
 		[fsw] "m" (tsk->thread.i387.fxsave.swd) : "memory");
 	/* AMD K7/K8 CPUs don't save/restore FDP/FIP/FOP unless an exception
 	   is pending.  Clear the x87 state here by setting it to fixed
-   	   values. __per_cpu_offset[0] is a random variable that should be in L1 */
+   	   values. safe_address is a random variable that should be in L1 */
 	alternative_input(
 		GENERIC_NOP8 GENERIC_NOP2,
 		"emms\n\t"	  	/* clear stack tags */
-- 
Chuck
"Penguins don't come from next door, they come from the Antarctic!"
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/