Re: [PATCH v2] sparc64: increase kernel thread stack size to 32K

[Imre Kaloz]

Hi Stian,

I ran this on a Sun Ultra 45 (UltraSPARC IIIi, SMP). The unpatched
kernel hits "corrupted stack end detected inside scheduler" in
hub_power_on() on 10 out of 10 boots with USB devices attached. With
the 32K stack it boots clean every time: 6 boots with the v1 diff,
10 more with the minimal diff below. A forced hub/device rebind with
the stack tracer armed peaked at 15832 of 32768 bytes, so 32K looks
like plenty here.

On v1, Andreas agreed with 32K but asked for the dead else branches
to be pulled into their own patch with a fixup tag for 15b9350a177b,
since that commit should have removed them already. v2 folds the
removal into the same patch, so I'd split it back out for v3: the
fix as the minimal diff below, the else-branch removal as its own
patch on top. Nice work!

The minimal diff is below the scissors, feel free to fold it in.

Tested-by: Imre Kaloz <kaloz@xxxxxxxxxx> # Sun Ultra 45


Best,
Imre

-- >8 --
Signed-off-by: Imre Kaloz <kaloz@xxxxxxxxxx>
---
diff --git a/arch/sparc/include/asm/thread_info_64.h b/arch/sparc/include/asm/thread_info_64.h
index c8a73dff27f8..ec552513d63c 100644
--- a/arch/sparc/include/asm/thread_info_64.h
+++ b/arch/sparc/include/asm/thread_info_64.h
@@ -100,8 +100,8 @@ struct thread_info {
 #define	FAULT_CODE_BAD_RA	0x20	/* Bad RA for sun4v		   */
 
 #if PAGE_SHIFT == 13
-#define THREAD_SIZE (2*PAGE_SIZE)
-#define THREAD_SHIFT (PAGE_SHIFT + 1)
+#define THREAD_SIZE (4*PAGE_SIZE)
+#define THREAD_SHIFT (PAGE_SHIFT + 2)
 #else /* PAGE_SHIFT == 13 */
 #define THREAD_SIZE PAGE_SIZE
 #define THREAD_SHIFT PAGE_SHIFT
@@ -129,7 +129,7 @@ extern struct thread_info *current_thread_info(void);
 
 /* thread information allocation */
 #if PAGE_SHIFT == 13
-#define THREAD_SIZE_ORDER	1
+#define THREAD_SIZE_ORDER	2
 #else /* PAGE_SHIFT == 13 */
 #define THREAD_SIZE_ORDER	0
 #endif /* PAGE_SHIFT == 13 */