[v1] iommu/dma: Trace bounce buffer usage when mapping buffers

[PATCH 5.15.y] iommu/dma: Trace bounce buffer usage when mapping buffers

Posted by Isaac J. Manjarres 1 year, 11 months ago

When commit 82612d66d51d ("iommu: Allow the dma-iommu api to
use bounce buffers") was introduced, it did not add the logic
for tracing the bounce buffer usage from iommu_dma_map_page().

All of the users of swiotlb_tbl_map_single() trace their bounce
buffer usage, except iommu_dma_map_page(). This makes it difficult
to track SWIOTLB usage from that function. Thus, trace bounce buffer
usage from iommu_dma_map_page().

Fixes: 82612d66d51d ("iommu: Allow the dma-iommu api to use bounce buffers")
Cc: stable@vger.kernel.org # v5.15+
Cc: Tom Murphy <murphyt7@tcd.ie>
Cc: Lu Baolu <baolu.lu@linux.intel.com>
Cc: Saravana Kannan <saravanak@google.com>
Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
Link: https://lore.kernel.org/r/20231208234141.2356157-1-isaacmanjarres@google.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
(cherry picked from commit a63c357b9fd56ad5fe64616f5b22835252c6a76a)
Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
---
 drivers/iommu/dma-iommu.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 48c6f7ff4aef..8cd63e6ccd2c 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -25,6 +25,7 @@
 #include <linux/vmalloc.h>
 #include <linux/crash_dump.h>
 #include <linux/dma-direct.h>
+#include <trace/events/swiotlb.h>
 
 struct iommu_dma_msi_page {
 	struct list_head	list;
@@ -817,6 +818,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
 		void *padding_start;
 		size_t padding_size, aligned_size;
 
+		trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
+
 		aligned_size = iova_align(iovad, size);
 		phys = swiotlb_tbl_map_single(dev, phys, size, aligned_size,
 					      iova_mask(iovad), dir, attrs);
-- 
2.43.0.429.g432eaa2c6b-goog

Re: [PATCH 5.15.y] iommu/dma: Trace bounce buffer usage when mapping buffers

Posted by Bibek Kumar Patro 1 year, 6 months ago

On Mon, Jan 22, 2024 at 12:37:54PM -0800, Isaac J. Manjarres wrote:
> When commit 82612d66d51d ("iommu: Allow the dma-iommu api to
> use bounce buffers") was introduced, it did not add the logic
> for tracing the bounce buffer usage from iommu_dma_map_page().
> 
> All of the users of swiotlb_tbl_map_single() trace their bounce
> buffer usage, except iommu_dma_map_page(). This makes it difficult
> to track SWIOTLB usage from that function. Thus, trace bounce buffer
> usage from iommu_dma_map_page().
> 
> Fixes: 82612d66d51d ("iommu: Allow the dma-iommu api to use bounce buffers")
> Cc: stable@vger.kernel.org # v5.15+
> Cc: Tom Murphy <murphyt7@tcd.ie>
> Cc: Lu Baolu <baolu.lu@linux.intel.com>
> Cc: Saravana Kannan <saravanak@google.com>
> Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> Link: https://lore.kernel.org/r/20231208234141.2356157-1-isaacmanjarres@google.com
> Signed-off-by: Joerg Roedel <jroedel@suse.de>
> (cherry picked from commit a63c357b9fd56ad5fe64616f5b22835252c6a76a)
> Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> ---
>  drivers/iommu/dma-iommu.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> index 48c6f7ff4aef..8cd63e6ccd2c 100644
> --- a/drivers/iommu/dma-iommu.c
> +++ b/drivers/iommu/dma-iommu.c
> @@ -25,6 +25,7 @@
>  #include <linux/vmalloc.h>
>  #include <linux/crash_dump.h>
>  #include <linux/dma-direct.h>
> +#include <trace/events/swiotlb.h>
>  
>  struct iommu_dma_msi_page {
>  	struct list_head	list;
> @@ -817,6 +818,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
>  		void *padding_start;
>  		size_t padding_size, aligned_size;
>  
> +		trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
> +

Hi, this backported patch trying to access swiotlb_force variable is
causing a build conflict where CONFIG_SWIOTLB is not enabled.

In file included from kernel/drivers/iommu/dma-iommu.c:28:
kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
                 enum swiotlb_force swiotlb_force),
                      ^
kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
#define swiotlb_force SWIOTLB_NO_FORCE
                      ^
In file included from kernel/drivers/iommu/dma-iommu.c:28:
kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
#define swiotlb_force SWIOTLB_NO_FORCE
                      ^
In file included from kernel/drivers/iommu/dma-iommu.c:28:
kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
#define swiotlb_force SWIOTLB_NO_FORCE
                      ^
In file included from kernel/drivers/iommu/dma-iommu.c:28:
kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
#define swiotlb_force SWIOTLB_NO_FORCE
                      ^
kernel/drivers/iommu/dma-iommu.c:865:42: error: argument type 'enum SWIOTLB_NO_FORCE' is incomplete
                                       trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
                                                                              ^~~~~~~~~~~~~
kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
#define swiotlb_force SWIOTLB_NO_FORCE
                      ^~~~~~~~~~~~~~~~
kernel/include/trace/events/swiotlb.h:15:9: note: forward declaration of 'enum SWIOTLB_NO_FORCE'
enum swiotlb_force swiotlb_force),
     ^
kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
#define swiotlb_force SWIOTLB_NO_FORCE

--------------------------------------------------------------------------------------------------------------------------------------------------

I have a simple proposed fix which can resolve this compile time conflict when CONFIG_SWIOTLB is disabled.

--- a/include/trace/events/swiotlb.h
+++ b/include/trace/events/swiotlb.h
@@ -7,6 +7,7 @@

 #include <linux/tracepoint.h>

+#ifdef CONFIG_SWIOTLB
 TRACE_EVENT(swiotlb_bounced,

        TP_PROTO(struct device *dev,
@@ -43,6 +44,9 @@ TRACE_EVENT(swiotlb_bounced,
                        { SWIOTLB_FORCE,        "FORCE" },
                        { SWIOTLB_NO_FORCE,     "NO_FORCE" }))
 );
+#else
+#define trace_swiotlb_bounced(dev, phys, size, swiotlb_force)
+#endif /* CONFIG_SWIOTLB */

 #endif /*  _TRACE_SWIOTLB_H */


Thanks & regards,
Bibek

>  		aligned_size = iova_align(iovad, size);
>  		phys = swiotlb_tbl_map_single(dev, phys, size, aligned_size,
>  					      iova_mask(iovad), dir, attrs);
> -- 
> 2.43.0.429.g432eaa2c6b-goog
>

Re: [PATCH 5.15.y] iommu/dma: Trace bounce buffer usage when mapping buffers

Posted by Greg KH 1 year, 6 months ago

On Thu, Jun 13, 2024 at 04:01:01PM +0530, Bibek Kumar Patro wrote:
> On Mon, Jan 22, 2024 at 12:37:54PM -0800, Isaac J. Manjarres wrote:
> > When commit 82612d66d51d ("iommu: Allow the dma-iommu api to
> > use bounce buffers") was introduced, it did not add the logic
> > for tracing the bounce buffer usage from iommu_dma_map_page().
> > 
> > All of the users of swiotlb_tbl_map_single() trace their bounce
> > buffer usage, except iommu_dma_map_page(). This makes it difficult
> > to track SWIOTLB usage from that function. Thus, trace bounce buffer
> > usage from iommu_dma_map_page().
> > 
> > Fixes: 82612d66d51d ("iommu: Allow the dma-iommu api to use bounce buffers")
> > Cc: stable@vger.kernel.org # v5.15+
> > Cc: Tom Murphy <murphyt7@tcd.ie>
> > Cc: Lu Baolu <baolu.lu@linux.intel.com>
> > Cc: Saravana Kannan <saravanak@google.com>
> > Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> > Link: https://lore.kernel.org/r/20231208234141.2356157-1-isaacmanjarres@google.com
> > Signed-off-by: Joerg Roedel <jroedel@suse.de>
> > (cherry picked from commit a63c357b9fd56ad5fe64616f5b22835252c6a76a)
> > Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> > ---
> >  drivers/iommu/dma-iommu.c | 3 +++
> >  1 file changed, 3 insertions(+)
> > 
> > diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> > index 48c6f7ff4aef..8cd63e6ccd2c 100644
> > --- a/drivers/iommu/dma-iommu.c
> > +++ b/drivers/iommu/dma-iommu.c
> > @@ -25,6 +25,7 @@
> >  #include <linux/vmalloc.h>
> >  #include <linux/crash_dump.h>
> >  #include <linux/dma-direct.h>
> > +#include <trace/events/swiotlb.h>
> >  
> >  struct iommu_dma_msi_page {
> >  	struct list_head	list;
> > @@ -817,6 +818,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> >  		void *padding_start;
> >  		size_t padding_size, aligned_size;
> >  
> > +		trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
> > +
> 
> Hi, this backported patch trying to access swiotlb_force variable is
> causing a build conflict where CONFIG_SWIOTLB is not enabled.
> 
> In file included from kernel/drivers/iommu/dma-iommu.c:28:
> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
>                  enum swiotlb_force swiotlb_force),
>                       ^
> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> #define swiotlb_force SWIOTLB_NO_FORCE
>                       ^
> In file included from kernel/drivers/iommu/dma-iommu.c:28:
> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> #define swiotlb_force SWIOTLB_NO_FORCE
>                       ^
> In file included from kernel/drivers/iommu/dma-iommu.c:28:
> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> #define swiotlb_force SWIOTLB_NO_FORCE
>                       ^
> In file included from kernel/drivers/iommu/dma-iommu.c:28:
> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> #define swiotlb_force SWIOTLB_NO_FORCE
>                       ^
> kernel/drivers/iommu/dma-iommu.c:865:42: error: argument type 'enum SWIOTLB_NO_FORCE' is incomplete
>                                        trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
>                                                                               ^~~~~~~~~~~~~
> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> #define swiotlb_force SWIOTLB_NO_FORCE
>                       ^~~~~~~~~~~~~~~~
> kernel/include/trace/events/swiotlb.h:15:9: note: forward declaration of 'enum SWIOTLB_NO_FORCE'
> enum swiotlb_force swiotlb_force),
>      ^
> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> #define swiotlb_force SWIOTLB_NO_FORCE
> 
> --------------------------------------------------------------------------------------------------------------------------------------------------
> 
> I have a simple proposed fix which can resolve this compile time conflict when CONFIG_SWIOTLB is disabled.
> 
> --- a/include/trace/events/swiotlb.h
> +++ b/include/trace/events/swiotlb.h
> @@ -7,6 +7,7 @@
> 
>  #include <linux/tracepoint.h>
> 
> +#ifdef CONFIG_SWIOTLB
>  TRACE_EVENT(swiotlb_bounced,
> 
>         TP_PROTO(struct device *dev,
> @@ -43,6 +44,9 @@ TRACE_EVENT(swiotlb_bounced,
>                         { SWIOTLB_FORCE,        "FORCE" },
>                         { SWIOTLB_NO_FORCE,     "NO_FORCE" }))
>  );
> +#else
> +#define trace_swiotlb_bounced(dev, phys, size, swiotlb_force)
> +#endif /* CONFIG_SWIOTLB */
> 
>  #endif /*  _TRACE_SWIOTLB_H */
> 
> 

Why not just take whatever change upstream fixes this instead of a
one-off change?

thanks

greg k-h

Re: [PATCH 5.15.y] iommu/dma: Trace bounce buffer usage when mapping buffers

Posted by Bibek Kumar Patro 1 year, 6 months ago


On 6/13/2024 4:45 PM, Greg KH wrote:
> On Thu, Jun 13, 2024 at 04:01:01PM +0530, Bibek Kumar Patro wrote:
>> On Mon, Jan 22, 2024 at 12:37:54PM -0800, Isaac J. Manjarres wrote:
>>> When commit 82612d66d51d ("iommu: Allow the dma-iommu api to
>>> use bounce buffers") was introduced, it did not add the logic
>>> for tracing the bounce buffer usage from iommu_dma_map_page().
>>>
>>> All of the users of swiotlb_tbl_map_single() trace their bounce
>>> buffer usage, except iommu_dma_map_page(). This makes it difficult
>>> to track SWIOTLB usage from that function. Thus, trace bounce buffer
>>> usage from iommu_dma_map_page().
>>>
>>> Fixes: 82612d66d51d ("iommu: Allow the dma-iommu api to use bounce buffers")
>>> Cc: stable@vger.kernel.org # v5.15+
>>> Cc: Tom Murphy <murphyt7@tcd.ie>
>>> Cc: Lu Baolu <baolu.lu@linux.intel.com>
>>> Cc: Saravana Kannan <saravanak@google.com>
>>> Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
>>> Link: https://lore.kernel.org/r/20231208234141.2356157-1-isaacmanjarres@google.com
>>> Signed-off-by: Joerg Roedel <jroedel@suse.de>
>>> (cherry picked from commit a63c357b9fd56ad5fe64616f5b22835252c6a76a)
>>> Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
>>> ---
>>>   drivers/iommu/dma-iommu.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
>>> index 48c6f7ff4aef..8cd63e6ccd2c 100644
>>> --- a/drivers/iommu/dma-iommu.c
>>> +++ b/drivers/iommu/dma-iommu.c
>>> @@ -25,6 +25,7 @@
>>>   #include <linux/vmalloc.h>
>>>   #include <linux/crash_dump.h>
>>>   #include <linux/dma-direct.h>
>>> +#include <trace/events/swiotlb.h>
>>>   
>>>   struct iommu_dma_msi_page {
>>>   	struct list_head	list;
>>> @@ -817,6 +818,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
>>>   		void *padding_start;
>>>   		size_t padding_size, aligned_size;
>>>   
>>> +		trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
>>> +
>>
>> Hi, this backported patch trying to access swiotlb_force variable is
>> causing a build conflict where CONFIG_SWIOTLB is not enabled.
>>
>> In file included from kernel/drivers/iommu/dma-iommu.c:28:
>> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
>>                   enum swiotlb_force swiotlb_force),
>>                        ^
>> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
>> #define swiotlb_force SWIOTLB_NO_FORCE
>>                        ^
>> In file included from kernel/drivers/iommu/dma-iommu.c:28:
>> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
>> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
>> #define swiotlb_force SWIOTLB_NO_FORCE
>>                        ^
>> In file included from kernel/drivers/iommu/dma-iommu.c:28:
>> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
>> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
>> #define swiotlb_force SWIOTLB_NO_FORCE
>>                        ^
>> In file included from kernel/drivers/iommu/dma-iommu.c:28:
>> kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
>> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
>> #define swiotlb_force SWIOTLB_NO_FORCE
>>                        ^
>> kernel/drivers/iommu/dma-iommu.c:865:42: error: argument type 'enum SWIOTLB_NO_FORCE' is incomplete
>>                                         trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
>>                                                                                ^~~~~~~~~~~~~
>> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
>> #define swiotlb_force SWIOTLB_NO_FORCE
>>                        ^~~~~~~~~~~~~~~~
>> kernel/include/trace/events/swiotlb.h:15:9: note: forward declaration of 'enum SWIOTLB_NO_FORCE'
>> enum swiotlb_force swiotlb_force),
>>       ^
>> kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
>> #define swiotlb_force SWIOTLB_NO_FORCE
>>
>> --------------------------------------------------------------------------------------------------------------------------------------------------
>>
>> I have a simple proposed fix which can resolve this compile time conflict when CONFIG_SWIOTLB is disabled.
>>
>> --- a/include/trace/events/swiotlb.h
>> +++ b/include/trace/events/swiotlb.h
>> @@ -7,6 +7,7 @@
>>
>>   #include <linux/tracepoint.h>
>>
>> +#ifdef CONFIG_SWIOTLB
>>   TRACE_EVENT(swiotlb_bounced,
>>
>>          TP_PROTO(struct device *dev,
>> @@ -43,6 +44,9 @@ TRACE_EVENT(swiotlb_bounced,
>>                          { SWIOTLB_FORCE,        "FORCE" },
>>                          { SWIOTLB_NO_FORCE,     "NO_FORCE" }))
>>   );
>> +#else
>> +#define trace_swiotlb_bounced(dev, phys, size, swiotlb_force)
>> +#endif /* CONFIG_SWIOTLB */
>>
>>   #endif /*  _TRACE_SWIOTLB_H */
>>
>>
> 
> Why not just take whatever change upstream fixes this instead of a
> one-off change?
> 

I am currently checking the history on swiotlb_force and how it's
removed in latest kernel versions. If those changes are applicable on
this stable branch can we explore backporting those instead of this one-
off change ?

Thanks & regards,
Bibek

> thanks
> 
> greg k-h

Re: [PATCH 5.15.y] iommu/dma: Trace bounce buffer usage when mapping buffers

Posted by Greg KH 1 year, 6 months ago

On Thu, Jun 13, 2024 at 11:10:57PM +0530, Bibek Kumar Patro wrote:
> 
> 
> On 6/13/2024 4:45 PM, Greg KH wrote:
> > On Thu, Jun 13, 2024 at 04:01:01PM +0530, Bibek Kumar Patro wrote:
> > > On Mon, Jan 22, 2024 at 12:37:54PM -0800, Isaac J. Manjarres wrote:
> > > > When commit 82612d66d51d ("iommu: Allow the dma-iommu api to
> > > > use bounce buffers") was introduced, it did not add the logic
> > > > for tracing the bounce buffer usage from iommu_dma_map_page().
> > > > 
> > > > All of the users of swiotlb_tbl_map_single() trace their bounce
> > > > buffer usage, except iommu_dma_map_page(). This makes it difficult
> > > > to track SWIOTLB usage from that function. Thus, trace bounce buffer
> > > > usage from iommu_dma_map_page().
> > > > 
> > > > Fixes: 82612d66d51d ("iommu: Allow the dma-iommu api to use bounce buffers")
> > > > Cc: stable@vger.kernel.org # v5.15+
> > > > Cc: Tom Murphy <murphyt7@tcd.ie>
> > > > Cc: Lu Baolu <baolu.lu@linux.intel.com>
> > > > Cc: Saravana Kannan <saravanak@google.com>
> > > > Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> > > > Link: https://lore.kernel.org/r/20231208234141.2356157-1-isaacmanjarres@google.com
> > > > Signed-off-by: Joerg Roedel <jroedel@suse.de>
> > > > (cherry picked from commit a63c357b9fd56ad5fe64616f5b22835252c6a76a)
> > > > Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> > > > ---
> > > >   drivers/iommu/dma-iommu.c | 3 +++
> > > >   1 file changed, 3 insertions(+)
> > > > 
> > > > diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> > > > index 48c6f7ff4aef..8cd63e6ccd2c 100644
> > > > --- a/drivers/iommu/dma-iommu.c
> > > > +++ b/drivers/iommu/dma-iommu.c
> > > > @@ -25,6 +25,7 @@
> > > >   #include <linux/vmalloc.h>
> > > >   #include <linux/crash_dump.h>
> > > >   #include <linux/dma-direct.h>
> > > > +#include <trace/events/swiotlb.h>
> > > >   struct iommu_dma_msi_page {
> > > >   	struct list_head	list;
> > > > @@ -817,6 +818,8 @@ static dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> > > >   		void *padding_start;
> > > >   		size_t padding_size, aligned_size;
> > > > +		trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
> > > > +
> > > 
> > > Hi, this backported patch trying to access swiotlb_force variable is
> > > causing a build conflict where CONFIG_SWIOTLB is not enabled.
> > > 
> > > In file included from kernel/drivers/iommu/dma-iommu.c:28:
> > > kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> > >                   enum swiotlb_force swiotlb_force),
> > >                        ^
> > > kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> > > #define swiotlb_force SWIOTLB_NO_FORCE
> > >                        ^
> > > In file included from kernel/drivers/iommu/dma-iommu.c:28:
> > > kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> > > kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> > > #define swiotlb_force SWIOTLB_NO_FORCE
> > >                        ^
> > > In file included from kernel/drivers/iommu/dma-iommu.c:28:
> > > kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> > > kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> > > #define swiotlb_force SWIOTLB_NO_FORCE
> > >                        ^
> > > In file included from kernel/drivers/iommu/dma-iommu.c:28:
> > > kernel/include/trace/events/swiotlb.h:15:9: error: declaration of 'enum SWIOTLB_NO_FORCE' will not be visible outside of this function [-Werror,-Wvisibility]
> > > kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> > > #define swiotlb_force SWIOTLB_NO_FORCE
> > >                        ^
> > > kernel/drivers/iommu/dma-iommu.c:865:42: error: argument type 'enum SWIOTLB_NO_FORCE' is incomplete
> > >                                         trace_swiotlb_bounced(dev, phys, size, swiotlb_force);
> > >                                                                                ^~~~~~~~~~~~~
> > > kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> > > #define swiotlb_force SWIOTLB_NO_FORCE
> > >                        ^~~~~~~~~~~~~~~~
> > > kernel/include/trace/events/swiotlb.h:15:9: note: forward declaration of 'enum SWIOTLB_NO_FORCE'
> > > enum swiotlb_force swiotlb_force),
> > >       ^
> > > kernel/include/linux/swiotlb.h:143:23: note: expanded from macro 'swiotlb_force'
> > > #define swiotlb_force SWIOTLB_NO_FORCE
> > > 
> > > --------------------------------------------------------------------------------------------------------------------------------------------------
> > > 
> > > I have a simple proposed fix which can resolve this compile time conflict when CONFIG_SWIOTLB is disabled.
> > > 
> > > --- a/include/trace/events/swiotlb.h
> > > +++ b/include/trace/events/swiotlb.h
> > > @@ -7,6 +7,7 @@
> > > 
> > >   #include <linux/tracepoint.h>
> > > 
> > > +#ifdef CONFIG_SWIOTLB
> > >   TRACE_EVENT(swiotlb_bounced,
> > > 
> > >          TP_PROTO(struct device *dev,
> > > @@ -43,6 +44,9 @@ TRACE_EVENT(swiotlb_bounced,
> > >                          { SWIOTLB_FORCE,        "FORCE" },
> > >                          { SWIOTLB_NO_FORCE,     "NO_FORCE" }))
> > >   );
> > > +#else
> > > +#define trace_swiotlb_bounced(dev, phys, size, swiotlb_force)
> > > +#endif /* CONFIG_SWIOTLB */
> > > 
> > >   #endif /*  _TRACE_SWIOTLB_H */
> > > 
> > > 
> > 
> > Why not just take whatever change upstream fixes this instead of a
> > one-off change?
> > 
> 
> I am currently checking the history on swiotlb_force and how it's
> removed in latest kernel versions. If those changes are applicable on
> this stable branch can we explore backporting those instead of this one-
> off change ?

Please backport what is in the tree already.

Re: [PATCH 5.15.y] iommu/dma: Trace bounce buffer usage when mapping buffers

Posted by Greg KH 1 year, 11 months ago

On Mon, Jan 22, 2024 at 12:37:54PM -0800, Isaac J. Manjarres wrote:
> When commit 82612d66d51d ("iommu: Allow the dma-iommu api to
> use bounce buffers") was introduced, it did not add the logic
> for tracing the bounce buffer usage from iommu_dma_map_page().
> 
> All of the users of swiotlb_tbl_map_single() trace their bounce
> buffer usage, except iommu_dma_map_page(). This makes it difficult
> to track SWIOTLB usage from that function. Thus, trace bounce buffer
> usage from iommu_dma_map_page().
> 
> Fixes: 82612d66d51d ("iommu: Allow the dma-iommu api to use bounce buffers")
> Cc: stable@vger.kernel.org # v5.15+
> Cc: Tom Murphy <murphyt7@tcd.ie>
> Cc: Lu Baolu <baolu.lu@linux.intel.com>
> Cc: Saravana Kannan <saravanak@google.com>
> Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>
> Link: https://lore.kernel.org/r/20231208234141.2356157-1-isaacmanjarres@google.com
> Signed-off-by: Joerg Roedel <jroedel@suse.de>
> (cherry picked from commit a63c357b9fd56ad5fe64616f5b22835252c6a76a)
> Signed-off-by: Isaac J. Manjarres <isaacmanjarres@google.com>

Now queued up, thanks.

greg k-h