* [PATCH -tip 0/3] dma-debug fixes
@ 2009-05-27 0:43 FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 1/3] dma-debug: use sg_dma_address accessor instead of using dma_address directly FUJITA Tomonori
` (3 more replies)
0 siblings, 4 replies; 6+ messages in thread
From: FUJITA Tomonori @ 2009-05-27 0:43 UTC (permalink / raw)
To: mingo; +Cc: joerg.roedel, fujita.tomonori, linux-kernel
None of them is critical (we get bogus warnings due to the bugs but
the bugs are in unpopular DMA mapping functions).
This patchset is against to tip/master.
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH -tip 1/3] dma-debug: use sg_dma_address accessor instead of using dma_address directly
2009-05-27 0:43 [PATCH -tip 0/3] dma-debug fixes FUJITA Tomonori
@ 2009-05-27 0:43 ` FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 2/3] dma-debug: use sg_dma_len accessor FUJITA Tomonori
` (2 subsequent siblings)
3 siblings, 0 replies; 6+ messages in thread
From: FUJITA Tomonori @ 2009-05-27 0:43 UTC (permalink / raw)
To: mingo; +Cc: joerg.roedel, fujita.tomonori, linux-kernel
Architectures might not have dma_address in struct scatterlist (PARISC
doesn't). Directly accessing to dma_address in struct scatterlist is
wrong; we need to use sg_dma_address() accesssor instead.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
lib/dma-debug.c | 12 ++++++------
1 files changed, 6 insertions(+), 6 deletions(-)
diff --git a/lib/dma-debug.c b/lib/dma-debug.c
index cdd205d..64aaa96 100644
--- a/lib/dma-debug.c
+++ b/lib/dma-debug.c
@@ -784,7 +784,7 @@ void debug_dma_map_sg(struct device *dev, struct scatterlist *sg,
entry->dev = dev;
entry->paddr = sg_phys(s);
entry->size = s->length;
- entry->dev_addr = s->dma_address;
+ entry->dev_addr = sg_dma_address(s);
entry->direction = direction;
entry->sg_call_ents = nents;
entry->sg_mapped_ents = mapped_ents;
@@ -816,7 +816,7 @@ void debug_dma_unmap_sg(struct device *dev, struct scatterlist *sglist,
.type = dma_debug_sg,
.dev = dev,
.paddr = sg_phys(s),
- .dev_addr = s->dma_address,
+ .dev_addr = sg_dma_address(s),
.size = s->length,
.direction = dir,
.sg_call_ents = 0,
@@ -940,8 +940,8 @@ void debug_dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
return;
for_each_sg(sg, s, nelems, i) {
- check_sync(dev, s->dma_address, s->dma_length, 0,
- direction, true);
+ check_sync(dev, sg_dma_address(s), s->dma_length, 0,
+ direction, true);
}
}
EXPORT_SYMBOL(debug_dma_sync_sg_for_cpu);
@@ -956,8 +956,8 @@ void debug_dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
return;
for_each_sg(sg, s, nelems, i) {
- check_sync(dev, s->dma_address, s->dma_length, 0,
- direction, false);
+ check_sync(dev, sg_dma_address(s), s->dma_length, 0,
+ direction, false);
}
}
EXPORT_SYMBOL(debug_dma_sync_sg_for_device);
--
1.6.0.6
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH -tip 2/3] dma-debug: use sg_dma_len accessor
2009-05-27 0:43 [PATCH -tip 0/3] dma-debug fixes FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 1/3] dma-debug: use sg_dma_address accessor instead of using dma_address directly FUJITA Tomonori
@ 2009-05-27 0:43 ` FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 3/3] dma-debug: fix debug_dma_sync_sg_for_cpu and debug_dma_sync_sg_for_device FUJITA Tomonori
2009-05-27 12:47 ` [PATCH -tip 0/3] dma-debug fixes Joerg Roedel
3 siblings, 0 replies; 6+ messages in thread
From: FUJITA Tomonori @ 2009-05-27 0:43 UTC (permalink / raw)
To: mingo; +Cc: joerg.roedel, fujita.tomonori, linux-kernel
debug_dma_map_sg() and debug_dma_unmap_sg() use length in struct
scatterlist while debug_dma_sync_sg_for_cpu() and
debug_dma_sync_sg_for_device() use dma_length. This causes bugs
warnings on some IOMMU implementations since these values are not
same; the length doesn't represent the dma length.
We always need to use sg_dma_len() accessor to get the dma length of a
scatterlist entry.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
lib/dma-debug.c | 10 +++++-----
1 files changed, 5 insertions(+), 5 deletions(-)
diff --git a/lib/dma-debug.c b/lib/dma-debug.c
index 64aaa96..df9e475 100644
--- a/lib/dma-debug.c
+++ b/lib/dma-debug.c
@@ -783,7 +783,7 @@ void debug_dma_map_sg(struct device *dev, struct scatterlist *sg,
entry->type = dma_debug_sg;
entry->dev = dev;
entry->paddr = sg_phys(s);
- entry->size = s->length;
+ entry->size = sg_dma_len(s);
entry->dev_addr = sg_dma_address(s);
entry->direction = direction;
entry->sg_call_ents = nents;
@@ -791,7 +791,7 @@ void debug_dma_map_sg(struct device *dev, struct scatterlist *sg,
if (!PageHighMem(sg_page(s))) {
check_for_stack(dev, sg_virt(s));
- check_for_illegal_area(dev, sg_virt(s), s->length);
+ check_for_illegal_area(dev, sg_virt(s), sg_dma_len(s));
}
add_dma_entry(entry);
@@ -817,7 +817,7 @@ void debug_dma_unmap_sg(struct device *dev, struct scatterlist *sglist,
.dev = dev,
.paddr = sg_phys(s),
.dev_addr = sg_dma_address(s),
- .size = s->length,
+ .size = sg_dma_len(s),
.direction = dir,
.sg_call_ents = 0,
};
@@ -940,7 +940,7 @@ void debug_dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
return;
for_each_sg(sg, s, nelems, i) {
- check_sync(dev, sg_dma_address(s), s->dma_length, 0,
+ check_sync(dev, sg_dma_address(s), sg_dma_len(s), 0,
direction, true);
}
}
@@ -956,7 +956,7 @@ void debug_dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
return;
for_each_sg(sg, s, nelems, i) {
- check_sync(dev, sg_dma_address(s), s->dma_length, 0,
+ check_sync(dev, sg_dma_address(s), sg_dma_len(s), 0,
direction, false);
}
}
--
1.6.0.6
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH -tip 3/3] dma-debug: fix debug_dma_sync_sg_for_cpu and debug_dma_sync_sg_for_device
2009-05-27 0:43 [PATCH -tip 0/3] dma-debug fixes FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 1/3] dma-debug: use sg_dma_address accessor instead of using dma_address directly FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 2/3] dma-debug: use sg_dma_len accessor FUJITA Tomonori
@ 2009-05-27 0:43 ` FUJITA Tomonori
2009-05-27 12:47 ` [PATCH -tip 0/3] dma-debug fixes Joerg Roedel
3 siblings, 0 replies; 6+ messages in thread
From: FUJITA Tomonori @ 2009-05-27 0:43 UTC (permalink / raw)
To: mingo; +Cc: joerg.roedel, fujita.tomonori, linux-kernel
DMA-mapping.txt says that debug_dma_sync_sg family must be called with
the _same_ one you passed into the dma_map_sg call, it should _NOT_ be
the 'count' value _returned_ from the dma_map_sg call.
debug_dma_sync_sg_for_cpu and debug_dma_sync_sg_for_device can't
handle this properly; they need to use the sg_mapped_ents in struct
dma_debug_entry as debug_dma_unmap_sg() does.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
lib/dma-debug.c | 48 +++++++++++++++++++++++++++++++++++++-----------
1 files changed, 37 insertions(+), 11 deletions(-)
diff --git a/lib/dma-debug.c b/lib/dma-debug.c
index df9e475..e10a981 100644
--- a/lib/dma-debug.c
+++ b/lib/dma-debug.c
@@ -799,13 +799,32 @@ void debug_dma_map_sg(struct device *dev, struct scatterlist *sg,
}
EXPORT_SYMBOL(debug_dma_map_sg);
+static int get_nr_mapped_entries(struct device *dev, struct scatterlist *s)
+{
+ struct dma_debug_entry *entry;
+ struct hash_bucket *bucket;
+ unsigned long flags;
+ int mapped_ents = 0;
+ struct dma_debug_entry ref;
+
+ ref.dev = dev;
+ ref.dev_addr = sg_dma_address(s);
+ ref.size = sg_dma_len(s),
+
+ bucket = get_hash_bucket(&ref, &flags);
+ entry = hash_bucket_find(bucket, &ref);
+ if (entry)
+ mapped_ents = entry->sg_mapped_ents;
+ put_hash_bucket(bucket, &flags);
+
+ return mapped_ents;
+}
+
void debug_dma_unmap_sg(struct device *dev, struct scatterlist *sglist,
int nelems, int dir)
{
- struct dma_debug_entry *entry;
struct scatterlist *s;
int mapped_ents = 0, i;
- unsigned long flags;
if (unlikely(global_disable))
return;
@@ -825,14 +844,9 @@ void debug_dma_unmap_sg(struct device *dev, struct scatterlist *sglist,
if (mapped_ents && i >= mapped_ents)
break;
- if (mapped_ents == 0) {
- struct hash_bucket *bucket;
+ if (!i) {
ref.sg_call_ents = nelems;
- bucket = get_hash_bucket(&ref, &flags);
- entry = hash_bucket_find(bucket, &ref);
- if (entry)
- mapped_ents = entry->sg_mapped_ents;
- put_hash_bucket(bucket, &flags);
+ mapped_ents = get_nr_mapped_entries(dev, s);
}
check_unmap(&ref);
@@ -934,12 +948,18 @@ void debug_dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
int nelems, int direction)
{
struct scatterlist *s;
- int i;
+ int mapped_ents = 0, i;
if (unlikely(global_disable))
return;
for_each_sg(sg, s, nelems, i) {
+ if (!i)
+ mapped_ents = get_nr_mapped_entries(dev, s);
+
+ if (i >= mapped_ents)
+ break;
+
check_sync(dev, sg_dma_address(s), sg_dma_len(s), 0,
direction, true);
}
@@ -950,12 +970,18 @@ void debug_dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
int nelems, int direction)
{
struct scatterlist *s;
- int i;
+ int mapped_ents = 0, i;
if (unlikely(global_disable))
return;
for_each_sg(sg, s, nelems, i) {
+ if (!i)
+ mapped_ents = get_nr_mapped_entries(dev, s);
+
+ if (i >= mapped_ents)
+ break;
+
check_sync(dev, sg_dma_address(s), sg_dma_len(s), 0,
direction, false);
}
--
1.6.0.6
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH -tip 0/3] dma-debug fixes
2009-05-27 0:43 [PATCH -tip 0/3] dma-debug fixes FUJITA Tomonori
` (2 preceding siblings ...)
2009-05-27 0:43 ` [PATCH -tip 3/3] dma-debug: fix debug_dma_sync_sg_for_cpu and debug_dma_sync_sg_for_device FUJITA Tomonori
@ 2009-05-27 12:47 ` Joerg Roedel
2009-05-27 23:33 ` Ingo Molnar
3 siblings, 1 reply; 6+ messages in thread
From: Joerg Roedel @ 2009-05-27 12:47 UTC (permalink / raw)
To: FUJITA Tomonori; +Cc: mingo, linux-kernel
On Wed, May 27, 2009 at 09:43:00AM +0900, FUJITA Tomonori wrote:
> None of them is critical (we get bogus warnings due to the bugs but
> the bugs are in unpopular DMA mapping functions).
>
> This patchset is against to tip/master.
Acked-by: Joerg Roedel <joerg.roedel@amd.com>
They all look good to me.
Ingo, do you want to pick them up youself or should I include them into
my dma-debug pull-request to you for 2.6.31?
Joerg
--
| Advanced Micro Devices GmbH
Operating | Karl-Hammerschmidt-Str. 34, 85609 Dornach bei München
System |
Research | Geschäftsführer: Thomas M. McCoy, Giuliano Meroni
Center | Sitz: Dornach, Gemeinde Aschheim, Landkreis München
| Registergericht München, HRB Nr. 43632
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH -tip 0/3] dma-debug fixes
2009-05-27 12:47 ` [PATCH -tip 0/3] dma-debug fixes Joerg Roedel
@ 2009-05-27 23:33 ` Ingo Molnar
0 siblings, 0 replies; 6+ messages in thread
From: Ingo Molnar @ 2009-05-27 23:33 UTC (permalink / raw)
To: Joerg Roedel; +Cc: FUJITA Tomonori, linux-kernel
* Joerg Roedel <joerg.roedel@amd.com> wrote:
> On Wed, May 27, 2009 at 09:43:00AM +0900, FUJITA Tomonori wrote:
> > None of them is critical (we get bogus warnings due to the bugs but
> > the bugs are in unpopular DMA mapping functions).
> >
> > This patchset is against to tip/master.
>
> Acked-by: Joerg Roedel <joerg.roedel@amd.com>
>
> They all look good to me.
>
> Ingo, do you want to pick them up youself or should I include them
> into my dma-debug pull-request to you for 2.6.31?
Sure, feel free to include them.
Thanks,
Ingo
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2009-05-27 23:34 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-05-27 0:43 [PATCH -tip 0/3] dma-debug fixes FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 1/3] dma-debug: use sg_dma_address accessor instead of using dma_address directly FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 2/3] dma-debug: use sg_dma_len accessor FUJITA Tomonori
2009-05-27 0:43 ` [PATCH -tip 3/3] dma-debug: fix debug_dma_sync_sg_for_cpu and debug_dma_sync_sg_for_device FUJITA Tomonori
2009-05-27 12:47 ` [PATCH -tip 0/3] dma-debug fixes Joerg Roedel
2009-05-27 23:33 ` Ingo Molnar
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).