[PATCH] fix powerpc breakage in sg chaining code (again)

[PATCH] fix powerpc breakage in sg chaining code (again)

[Johannes Berg]

Commit 33ff910f0f466184ffc3514628f18403dcd86761 fixed commit
78bdc3106a877cfa50439fa66b52acbc4e7868df but the code that got put in
uses struct scatterlist->page which no longer exists since commit
18dabf473e15850c0dbc8ff13ac1e2806d542c15 which requires using
sg_page(sg) instead of sg->page.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
If it was fixed before please ignore, I just ran into this.

--- linux-2.6.orig/include/asm/dma-mapping.h	2007-10-23 22:55:33.551133142 +0200
+++ linux-2.6/include/asm/dma-mapping.h	2007-10-23 22:56:07.081133142 +0200
@@ -285,9 +285,9 @@ dma_map_sg(struct device *dev, struct sc
 	BUG_ON(direction == DMA_NONE);
 
 	for_each_sg(sgl, sg, nents, i) {
-		BUG_ON(!sg->page);
-		__dma_sync_page(sg->page, sg->offset, sg->length, direction);
-		sg->dma_address = page_to_bus(sg->page) + sg->offset;
+		BUG_ON(!sg_page(sg));
+		__dma_sync_page(sg_page(sg), sg->offset, sg->length, direction);
+		sg->dma_address = page_to_bus(sg_page(sg)) + sg->offset;
 	}
 
 	return nents;
@@ -328,7 +328,7 @@ static inline void dma_sync_sg_for_cpu(s
 	BUG_ON(direction == DMA_NONE);
 
 	for_each_sg(sgl, sg, nents, i)
-		__dma_sync_page(sg->page, sg->offset, sg->length, direction);
+		__dma_sync_page(sg_page(sg), sg->offset, sg->length, direction);
 }
 
 static inline void dma_sync_sg_for_device(struct device *dev,
@@ -341,7 +341,7 @@ static inline void dma_sync_sg_for_devic
 	BUG_ON(direction == DMA_NONE);
 
 	for_each_sg(sgl, sg, nents, i)
-		__dma_sync_page(sg->page, sg->offset, sg->length, direction);
+		__dma_sync_page(sg_page(sg), sg->offset, sg->length, direction);
 }
 
 static inline int dma_mapping_error(dma_addr_t dma_addr)

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 805 bytes --]

On Tue, 2007-10-23 at 23:00 +0200, Johannes Berg wrote:
> Commit 33ff910f0f466184ffc3514628f18403dcd86761 fixed commit
> 78bdc3106a877cfa50439fa66b52acbc4e7868df but the code that got put in
> uses struct scatterlist->page which no longer exists since commit
> 18dabf473e15850c0dbc8ff13ac1e2806d542c15 which requires using
> sg_page(sg) instead of sg->page.

This doesn't help though, when I boot I get a NULL-pointer dereference
that looks approximately like this:

NIP blk_rq_map_sg + 0x64/0x190
LR ide_map_sg + 0x38/?
Call trace:
cfq_remove_request + 0x168 (unreliable)
cfq_dispatch_request + 0x44
ide_build_sglist + 0x38?
pmac_ide_dma_setup + 0xa0?
ide_do_rw_disk? + ?
[...]

I can upload the picture I took but it's hardly readable (taken with my
crappy cell phone)

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Jens Axboe]

On Tue, Oct 23 2007, Johannes Berg wrote:
> On Tue, 2007-10-23 at 23:00 +0200, Johannes Berg wrote:
> > Commit 33ff910f0f466184ffc3514628f18403dcd86761 fixed commit
> > 78bdc3106a877cfa50439fa66b52acbc4e7868df but the code that got put in
> > uses struct scatterlist->page which no longer exists since commit
> > 18dabf473e15850c0dbc8ff13ac1e2806d542c15 which requires using
> > sg_page(sg) instead of sg->page.
> 
> This doesn't help though, when I boot I get a NULL-pointer dereference
> that looks approximately like this:
> 
> NIP blk_rq_map_sg + 0x64/0x190
> LR ide_map_sg + 0x38/?
> Call trace:
> cfq_remove_request + 0x168 (unreliable)
> cfq_dispatch_request + 0x44
> ide_build_sglist + 0x38?
> pmac_ide_dma_setup + 0xa0?
> ide_do_rw_disk? + ?
> [...]
> 
> I can upload the picture I took but it's hardly readable (taken with my
> crappy cell phone)

I lost track - which kernel are you booting? This looks like something
that should be fixed, did you try 2.6.24-rc1?

-- 
Jens Axboe

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 431 bytes --]

On Wed, 2007-10-24 at 11:14 +0200, Jens Axboe wrote:

> I lost track - which kernel are you booting? This looks like something
> that should be fixed, did you try 2.6.24-rc1?

No, it came out after I pulled, I was on v2.6.23-6815-g0895e91. I'll
give it a try, but I don't think I can confirm it works before tomorrow.
I see the build failure got fixed with commit
5edadbd0ae35d2daabaf6b44f2c58d67d4021ed2 too.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Jens Axboe]

On Wed, Oct 24 2007, Johannes Berg wrote:
> On Wed, 2007-10-24 at 11:14 +0200, Jens Axboe wrote:
> 
> > I lost track - which kernel are you booting? This looks like something
> > that should be fixed, did you try 2.6.24-rc1?
> 
> No, it came out after I pulled, I was on v2.6.23-6815-g0895e91. I'll
> give it a try, but I don't think I can confirm it works before tomorrow.
> I see the build failure got fixed with commit
> 5edadbd0ae35d2daabaf6b44f2c58d67d4021ed2 too.

0895e91d60ef9bdef426d1ce14bb94bd5875870d is definitely too old, so it
will break on IDE. I'm confident that a newer kernel will solve this
issue.

-- 
Jens Axboe

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 437 bytes --]

On Wed, 2007-10-24 at 11:22 +0200, Johannes Berg wrote:

> No, it came out after I pulled, I was on v2.6.23-6815-g0895e91. I'll
> give it a try, but I don't think I can confirm it works before tomorrow.
> I see the build failure got fixed with commit
> 5edadbd0ae35d2daabaf6b44f2c58d67d4021ed2 too.

Wait, now I lost track. This patch is identical to that commit 5edad...,
what I was thinking of was the oops I got.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 619 bytes --]

On Wed, 2007-10-24 at 11:23 +0200, Johannes Berg wrote:
> On Wed, 2007-10-24 at 11:22 +0200, Johannes Berg wrote:
> 
> > No, it came out after I pulled, I was on v2.6.23-6815-g0895e91. I'll
> > give it a try, but I don't think I can confirm it works before tomorrow.
> > I see the build failure got fixed with commit
> > 5edadbd0ae35d2daabaf6b44f2c58d67d4021ed2 too.
> 
> Wait, now I lost track. This patch is identical to that commit 5edad...,
> what I was thinking of was the oops I got.

Nah, never mind. Apologies to everybody. I'm confused, Jens says the
oops was fixed. I'll verify that.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]

Re: [PATCH] fix powerpc breakage in sg chaining code (again)

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 774 bytes --]

On Wed, 2007-10-24 at 11:23 +0200, Jens Axboe wrote:
> On Wed, Oct 24 2007, Johannes Berg wrote:
> > On Wed, 2007-10-24 at 11:14 +0200, Jens Axboe wrote:
> > 
> > > I lost track - which kernel are you booting? This looks like something
> > > that should be fixed, did you try 2.6.24-rc1?
> > 
> > No, it came out after I pulled, I was on v2.6.23-6815-g0895e91. I'll
> > give it a try, but I don't think I can confirm it works before tomorrow.
> > I see the build failure got fixed with commit
> > 5edadbd0ae35d2daabaf6b44f2c58d67d4021ed2 too.
> 
> 0895e91d60ef9bdef426d1ce14bb94bd5875870d is definitely too old, so it
> will break on IDE. I'm confident that a newer kernel will solve this
> issue.

It does indeed, 2.6.24-rc1 runs fine. Thanks.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]