[PATCH v2 0/2] dma-buf: system_heap: account for system heap allocation in memcg

[PATCH v2 0/2] dma-buf: system_heap: account for system heap allocation in memcg

[Eric Chanudet]

Capture dmabuf system heap allocations in memcg following prior
conversations[1][2]. Disable this behavior by default unless configured
by "dma_heap.mem_accounting" module parameter.

[1] https://lore.kernel.org/dri-devel/Z-5GZ3kJDbhgVBPG@phenom.ffwll.local/
[2] https://lore.kernel.org/all/CABdmKX2_UOENujpW0dXe0Z0x+4V3onfGDmHf1DMOXfDha6ddOA@mail.gmail.com/

Changes in v2:
- Add a module parameter to enable dma-buf cgroup accounting, disabled
  by default.
- Split system_heap logic in its own commit.
- Link to v1: https://lore.kernel.org/lkml/20251211193106.755485-2-echanude@redhat.com/

Signed-off-by: Eric Chanudet <echanude@redhat.com>
---
Eric Chanudet (2):
      dma-buf: heaps: add parameter to account allocations using cgroup
      dma-buf: system_heap: account for system heap allocation in memcg

 drivers/dma-buf/dma-heap.c          | 5 +++++
 drivers/dma-buf/heaps/system_heap.c | 9 +++++++--
 2 files changed, 12 insertions(+), 2 deletions(-)
---
base-commit: b71e635feefc852405b14620a7fc58c4c80c0f73
change-id: 20260102-dmabuf-heap-system-memcg-c86a381d663a

Best regards,
-- 
Eric Chanudet <echanude@redhat.com>

[PATCH v2 1/2] dma-buf: heaps: add parameter to account allocations using cgroup

[Eric Chanudet]

Add a parameter to enable dma-buf heaps allocation accounting using
cgroup for heaps that implement it. It is disabled by default as doing
so incurs caveats based on how memcg currently accounts for shared
buffers.

Signed-off-by: Eric Chanudet <echanude@redhat.com>
---
 drivers/dma-buf/dma-heap.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/dma-buf/dma-heap.c b/drivers/dma-buf/dma-heap.c
index 8ab49924f8b71a0272dc89a609539a429feaf6c8..f72e93cb8367fa196e5b87e762a60463023887df 100644
--- a/drivers/dma-buf/dma-heap.c
+++ b/drivers/dma-buf/dma-heap.c
@@ -49,6 +49,11 @@ static dev_t dma_heap_devt;
 static struct class *dma_heap_class;
 static DEFINE_XARRAY_ALLOC(dma_heap_minors);
 
+bool mem_accounting;
+module_param(mem_accounting, bool, 0444);
+MODULE_PARM_DESC(mem_accounting,
+		 "Enable cgroup-based memory accounting for dma-buf heap allocations (default=false).");
+
 static int dma_heap_buffer_alloc(struct dma_heap *heap, size_t len,
 				 u32 fd_flags,
 				 u64 heap_flags)

-- 
2.52.0

[PATCH v2 2/2] dma-buf: system_heap: account for system heap allocation in memcg

[Eric Chanudet]

The system dma-buf heap lets userspace allocate buffers from the page
allocator. However, these allocations are not accounted for in memcg,
allowing processes to escape limits that may be configured.

Pass __GFP_ACCOUNT for system heap allocations, based on the
dma_heap.mem_accounting parameter, to use memcg and account for them.

Signed-off-by: Eric Chanudet <echanude@redhat.com>
---
 drivers/dma-buf/heaps/system_heap.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/dma-buf/heaps/system_heap.c b/drivers/dma-buf/heaps/system_heap.c
index 4c782fe33fd497a74eb5065797259576f9b651b6..139b50df64ed4c4a6fdd69f25fe48324fbe2c481 100644
--- a/drivers/dma-buf/heaps/system_heap.c
+++ b/drivers/dma-buf/heaps/system_heap.c
@@ -52,6 +52,8 @@ static gfp_t order_flags[] = {HIGH_ORDER_GFP, HIGH_ORDER_GFP, LOW_ORDER_GFP};
 static const unsigned int orders[] = {8, 4, 0};
 #define NUM_ORDERS ARRAY_SIZE(orders)
 
+extern bool mem_accounting;
+
 static int dup_sg_table(struct sg_table *from, struct sg_table *to)
 {
 	struct scatterlist *sg, *new_sg;
@@ -320,14 +322,17 @@ static struct page *alloc_largest_available(unsigned long size,
 {
 	struct page *page;
 	int i;
+	gfp_t flags;
 
 	for (i = 0; i < NUM_ORDERS; i++) {
 		if (size <  (PAGE_SIZE << orders[i]))
 			continue;
 		if (max_order < orders[i])
 			continue;
-
-		page = alloc_pages(order_flags[i], orders[i]);
+		flags = order_flags[i];
+		if (mem_accounting)
+			flags |= __GFP_ACCOUNT;
+		page = alloc_pages(flags, orders[i]);
 		if (!page)
 			continue;
 		return page;

-- 
2.52.0

Re: [PATCH v2 2/2] dma-buf: system_heap: account for system heap allocation in memcg

[Christian König]

On 1/13/26 22:32, Eric Chanudet wrote:
> The system dma-buf heap lets userspace allocate buffers from the page
> allocator. However, these allocations are not accounted for in memcg,
> allowing processes to escape limits that may be configured.
> 
> Pass __GFP_ACCOUNT for system heap allocations, based on the
> dma_heap.mem_accounting parameter, to use memcg and account for them.
> 
> Signed-off-by: Eric Chanudet <echanude@redhat.com>
> ---
>  drivers/dma-buf/heaps/system_heap.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/dma-buf/heaps/system_heap.c b/drivers/dma-buf/heaps/system_heap.c
> index 4c782fe33fd497a74eb5065797259576f9b651b6..139b50df64ed4c4a6fdd69f25fe48324fbe2c481 100644
> --- a/drivers/dma-buf/heaps/system_heap.c
> +++ b/drivers/dma-buf/heaps/system_heap.c
> @@ -52,6 +52,8 @@ static gfp_t order_flags[] = {HIGH_ORDER_GFP, HIGH_ORDER_GFP, LOW_ORDER_GFP};
>  static const unsigned int orders[] = {8, 4, 0};
>  #define NUM_ORDERS ARRAY_SIZE(orders)
>  
> +extern bool mem_accounting;

Please define that in some header. Apart from that looks good technically.

But after the discussion it sounds more and more like we don't want to account device driver allocated memory in memcg at all.

Regards,
Christian.


> +
>  static int dup_sg_table(struct sg_table *from, struct sg_table *to)
>  {
>  	struct scatterlist *sg, *new_sg;
> @@ -320,14 +322,17 @@ static struct page *alloc_largest_available(unsigned long size,
>  {
>  	struct page *page;
>  	int i;
> +	gfp_t flags;
>  
>  	for (i = 0; i < NUM_ORDERS; i++) {
>  		if (size <  (PAGE_SIZE << orders[i]))
>  			continue;
>  		if (max_order < orders[i])
>  			continue;
> -
> -		page = alloc_pages(order_flags[i], orders[i]);
> +		flags = order_flags[i];
> +		if (mem_accounting)
> +			flags |= __GFP_ACCOUNT;
> +		page = alloc_pages(flags, orders[i]);
>  		if (!page)
>  			continue;
>  		return page;
> 

Re: [PATCH v2 2/2] dma-buf: system_heap: account for system heap allocation in memcg

[Eric Chanudet]

On Wed, Jan 14, 2026 at 11:38:27AM +0100, Christian König wrote:
> On 1/13/26 22:32, Eric Chanudet wrote:
> > The system dma-buf heap lets userspace allocate buffers from the page
> > allocator. However, these allocations are not accounted for in memcg,
> > allowing processes to escape limits that may be configured.
> > 
> > Pass __GFP_ACCOUNT for system heap allocations, based on the
> > dma_heap.mem_accounting parameter, to use memcg and account for them.
> > 
> > Signed-off-by: Eric Chanudet <echanude@redhat.com>
> > ---
> >  drivers/dma-buf/heaps/system_heap.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/dma-buf/heaps/system_heap.c b/drivers/dma-buf/heaps/system_heap.c
> > index 4c782fe33fd497a74eb5065797259576f9b651b6..139b50df64ed4c4a6fdd69f25fe48324fbe2c481 100644
> > --- a/drivers/dma-buf/heaps/system_heap.c
> > +++ b/drivers/dma-buf/heaps/system_heap.c
> > @@ -52,6 +52,8 @@ static gfp_t order_flags[] = {HIGH_ORDER_GFP, HIGH_ORDER_GFP, LOW_ORDER_GFP};
> >  static const unsigned int orders[] = {8, 4, 0};
> >  #define NUM_ORDERS ARRAY_SIZE(orders)
> >  
> > +extern bool mem_accounting;
> 
> Please define that in some header. Apart from that looks good technically.

Thank you for the review, I can move it to linux/dma-heap.h in a v3
since it's intended for other heaps as well.

> But after the discussion it sounds more and more like we don't want to account device driver allocated memory in memcg at all.

From the threads in v1 I thought adding the switch left open a
consideration to use memcg with driver allocated memory for userspace,
even with the known caveats that implies. Re-reading your last reply[1],
that's not quite the case it sounds like.

Best,

[1] https://lore.kernel.org/all/e38d87d3-a114-43f9-be93-03e9b9f40844@amd.com/

> 
> Regards,
> Christian.
> 
> 
> > +
> >  static int dup_sg_table(struct sg_table *from, struct sg_table *to)
> >  {
> >  	struct scatterlist *sg, *new_sg;
> > @@ -320,14 +322,17 @@ static struct page *alloc_largest_available(unsigned long size,
> >  {
> >  	struct page *page;
> >  	int i;
> > +	gfp_t flags;
> >  
> >  	for (i = 0; i < NUM_ORDERS; i++) {
> >  		if (size <  (PAGE_SIZE << orders[i]))
> >  			continue;
> >  		if (max_order < orders[i])
> >  			continue;
> > -
> > -		page = alloc_pages(order_flags[i], orders[i]);
> > +		flags = order_flags[i];
> > +		if (mem_accounting)
> > +			flags |= __GFP_ACCOUNT;
> > +		page = alloc_pages(flags, orders[i]);
> >  		if (!page)
> >  			continue;
> >  		return page;
> > 
> 

-- 
Eric Chanudet

Re: [PATCH v2 0/2] dma-buf: system_heap: account for system heap allocation in memcg

[T.J. Mercier]

On Tue, Jan 13, 2026 at 1:33 PM Eric Chanudet <echanude@redhat.com> wrote:
>
> Capture dmabuf system heap allocations in memcg following prior
> conversations[1][2]. Disable this behavior by default unless configured
> by "dma_heap.mem_accounting" module parameter.
>
> [1] https://lore.kernel.org/dri-devel/Z-5GZ3kJDbhgVBPG@phenom.ffwll.local/
> [2] https://lore.kernel.org/all/CABdmKX2_UOENujpW0dXe0Z0x+4V3onfGDmHf1DMOXfDha6ddOA@mail.gmail.com/
>
> Changes in v2:
> - Add a module parameter to enable dma-buf cgroup accounting, disabled
>   by default.
> - Split system_heap logic in its own commit.
> - Link to v1: https://lore.kernel.org/lkml/20251211193106.755485-2-echanude@redhat.com/
>
> Signed-off-by: Eric Chanudet <echanude@redhat.com>
> ---
> Eric Chanudet (2):
>       dma-buf: heaps: add parameter to account allocations using cgroup
>       dma-buf: system_heap: account for system heap allocation in memcg
>
>  drivers/dma-buf/dma-heap.c          | 5 +++++
>  drivers/dma-buf/heaps/system_heap.c | 9 +++++++--
>  2 files changed, 12 insertions(+), 2 deletions(-)
> ---
> base-commit: b71e635feefc852405b14620a7fc58c4c80c0f73
> change-id: 20260102-dmabuf-heap-system-memcg-c86a381d663a
>
> Best regards,
> --
> Eric Chanudet <echanude@redhat.com>

Thanks for adding the param, LGTM. Looking forward to v3.