On Sun, 14 Nov 2010, John David Anglin wrote:
> I've pretty much narrowed down where the unaligned load is. It coming from
> init_page_buffers. I had to disable inlining of this function to simplify
> the code.
>
> 00000000401f9698 <init_page_buffers>:
> 401f9698: 08 03 02 41 copy r3,r1
> 401f969c: 08 1e 02 43 copy sp,r3
> 401f96a0: 73 c1 00 88 std,ma r1,40(sp)
> 401f96a4: 0f 40 10 dc ldd 0(r26),ret0
> 401f96a8: c6 9c c0 e8 bb,>= ret0,14,401f9724 <init_page_buffers+0x8c>
> 401f96ac: db 5c 0b fd extrd,u r26,63,3,ret0
> 401f96b0: bf 80 20 e8 cmpb,*<> r0,ret0,401f972c <init_page_buffers+0x94>
> 401f96b4: 53 54 00 20 ldd 10(r26),r20
> 401f96b8: 0f 40 10 dc ldd 0(r26),ret0
> 401f96bc: db 9a 0b 9f extrd,u ret0,60,1,r26
> 401f96c0: 0b 40 28 a0 cmpclr,*= r0,r26,r0
> 401f96c4: 08 14 02 53 copy r20,r19
> -> 401f96c8: 0e 60 10 dc ldd 0(r19),ret0
This hpmc seems to be a gcc miscompilation:
ii gcc-4.3-hppa64 4.3.5-4 The GNU C compiler (cross compiler for hppa6
The hpmc is avoided by not inlining init_page_buffers().
Trying 4.4.5. I believe 4.3.5-4 was from the initial lenny install.
Dave
--
J. David Anglin dave.anglin@xxxxxxxxxxxxxx
National Research Council of Canada (613) 990-0752 (FAX: 952-6602)
diff --git a/fs/buffer.c b/fs/buffer.c
index 3e7dca2..654128e 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -963,7 +963,7 @@ link_dev_buffers(struct page *page, struct buffer_head *head)
/*
* Initialise the state of a blockdev page's buffers.
*/
-static void
+static __attribute__((__noinline__)) void
init_page_buffers(struct page *page, struct block_device *bdev,
sector_t block, int size)
{
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
