Linux kernel staging patches
 help / color / mirror / Atom feed
* [PATCH v3] staging: octeon: fix asymmetric output buffer pool accounting
@ 2026-11-16  6:57 Priyanshu Singh
  2026-11-16  9:40 ` Greg Kroah-Hartman
  0 siblings, 1 reply; 3+ messages in thread
From: Priyanshu Singh @ 2026-11-16  6:57 UTC (permalink / raw)
  To: Greg Kroah-Hartman
  Cc: Dan Carpenter, Oscar Alejandro Sánchez Lucero, linux-staging,
	linux-kernel, Priyanshu Singh

In cvm_oct_configure_common_hw(), when CVMX_FPA_OUTPUT_BUFFER_POOL differs
from CVMX_FPA_PACKET_POOL, 1024 output buffers are allocated and added to
the hardware Free Pool Allocator (FPA) pool. However, in cvm_oct_probe()
(err_tx:) and in cvm_oct_remove(), cvm_oct_mem_empty_fpa() is called with
128 instead of 1024 as the expected buffer count.

This asymmetric accounting causes cvm_oct_free_hw_memory() to report a
false warning ("Freeing of pool 2 had too many buffers (-896)") upon
draining the pool.

Define CVM_OCT_NUM_OUTPUT_BUFFERS in ethernet-defines.h and use this
constant across cvm_oct_configure_common_hw(),
cvm_oct_probe(), and cvm_oct_remove() to ensure symmetric allocation
and deallocation.

This was found via manual code inspection and is compile-tested only.

Fixes: d5f9bc7360fc ("STAGING: Octeon: Increase output command buffers")
Fixes: 4bcc688a11da ("staging: octeon: handle rx/tx initialization failures in probe")
Signed-off-by: Priyanshu Singh <singhpriyanshu9838@gmail.com>
---
v2 -> v3:
- Add Fixes tag.

v1 -> v2:
- Make it clear in commit message that this was found via code inspection
  and is compile-tested only.
- Fix missing space typo in commit message.

 drivers/staging/octeon/ethernet-defines.h | 2 ++
 drivers/staging/octeon/ethernet.c         | 9 ++++++---
 2 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/drivers/staging/octeon/ethernet-defines.h b/drivers/staging/octeon/ethernet-defines.h
index c060374a3da2..3ce7c125ccfd 100644
--- a/drivers/staging/octeon/ethernet-defines.h
+++ b/drivers/staging/octeon/ethernet-defines.h
@@ -37,4 +37,6 @@
 
 #define TOTAL_NUMBER_OF_PORTS		(CVMX_PIP_NUM_INPUT_PORTS + 1)
 
+#define CVM_OCT_NUM_OUTPUT_BUFFERS	1024
+
 #endif /* __ETHERNET_DEFINES_H__ */
diff --git a/drivers/staging/octeon/ethernet.c b/drivers/staging/octeon/ethernet.c
index dea5696a6642..e0e6d40602f0 100644
--- a/drivers/staging/octeon/ethernet.c
+++ b/drivers/staging/octeon/ethernet.c
@@ -162,7 +162,8 @@ static void cvm_oct_configure_common_hw(struct platform_device *pdev)
 			     num_packet_buffers);
 	if (CVMX_FPA_OUTPUT_BUFFER_POOL != CVMX_FPA_PACKET_POOL)
 		cvm_oct_mem_fill_fpa(pdev, CVMX_FPA_OUTPUT_BUFFER_POOL,
-				     CVMX_FPA_OUTPUT_BUFFER_POOL_SIZE, 1024);
+				     CVMX_FPA_OUTPUT_BUFFER_POOL_SIZE,
+				     CVM_OCT_NUM_OUTPUT_BUFFERS);
 
 #ifdef __LITTLE_ENDIAN
 	{
@@ -974,7 +975,8 @@ static int cvm_oct_probe(struct platform_device *pdev)
 			      num_packet_buffers);
 	if (CVMX_FPA_OUTPUT_BUFFER_POOL != CVMX_FPA_PACKET_POOL)
 		cvm_oct_mem_empty_fpa(pdev, CVMX_FPA_OUTPUT_BUFFER_POOL,
-				      CVMX_FPA_OUTPUT_BUFFER_POOL_SIZE, 128);
+				      CVMX_FPA_OUTPUT_BUFFER_POOL_SIZE,
+				      CVM_OCT_NUM_OUTPUT_BUFFERS);
 
 	return ret;
 }
@@ -1009,7 +1011,8 @@ static void cvm_oct_remove(struct platform_device *pdev)
 			      num_packet_buffers);
 	if (CVMX_FPA_OUTPUT_BUFFER_POOL != CVMX_FPA_PACKET_POOL)
 		cvm_oct_mem_empty_fpa(pdev, CVMX_FPA_OUTPUT_BUFFER_POOL,
-				      CVMX_FPA_OUTPUT_BUFFER_POOL_SIZE, 128);
+				      CVMX_FPA_OUTPUT_BUFFER_POOL_SIZE,
+				      CVM_OCT_NUM_OUTPUT_BUFFERS);
 }
 
 static const struct of_device_id cvm_oct_match[] = {
-- 
2.53.0


^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH v3] staging: octeon: fix asymmetric output buffer pool accounting
  2026-11-16  6:57 [PATCH v3] staging: octeon: fix asymmetric output buffer pool accounting Priyanshu Singh
@ 2026-11-16  9:40 ` Greg Kroah-Hartman
  2026-11-16 10:16   ` ASH
  0 siblings, 1 reply; 3+ messages in thread
From: Greg Kroah-Hartman @ 2026-11-16  9:40 UTC (permalink / raw)
  To: Priyanshu Singh
  Cc: Dan Carpenter, Oscar Alejandro Sánchez Lucero, linux-staging,
	linux-kernel

On Mon, Nov 16, 2026 at 06:57:30AM +0000, Priyanshu Singh wrote:
> In cvm_oct_configure_common_hw(), when CVMX_FPA_OUTPUT_BUFFER_POOL differs
> from CVMX_FPA_PACKET_POOL, 1024 output buffers are allocated and added to
> the hardware Free Pool Allocator (FPA) pool. However, in cvm_oct_probe()
> (err_tx:) and in cvm_oct_remove(), cvm_oct_mem_empty_fpa() is called with
> 128 instead of 1024 as the expected buffer count.
> 
> This asymmetric accounting causes cvm_oct_free_hw_memory() to report a
> false warning ("Freeing of pool 2 had too many buffers (-896)") upon
> draining the pool.
> 
> Define CVM_OCT_NUM_OUTPUT_BUFFERS in ethernet-defines.h and use this
> constant across cvm_oct_configure_common_hw(),
> cvm_oct_probe(), and cvm_oct_remove() to ensure symmetric allocation
> and deallocation.
> 
> This was found via manual code inspection and is compile-tested only.

Please test this on real hardware in order for us to be able to accept
it.

thanks,

greg k-h

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH v3] staging: octeon: fix asymmetric output buffer pool accounting
  2026-11-16  9:40 ` Greg Kroah-Hartman
@ 2026-11-16 10:16   ` ASH
  0 siblings, 0 replies; 3+ messages in thread
From: ASH @ 2026-11-16 10:16 UTC (permalink / raw)
  To: Greg Kroah-Hartman
  Cc: Dan Carpenter, Oscar Alejandro Sánchez Lucero, linux-staging,
	linux-kernel

On Mon, Nov 16, 2026 at 3:10 PM Greg Kroah-Hartman
<gregkh@linuxfoundation.org> wrote:
> Please test this on real hardware in order for us to be able to accept
> it.

Unfortunately, I do not have access to physical Octeon hardware to
run-test this.

The patch was based on commit d5f9bc7360fc ("STAGING: Octeon: Increase output
command buffers"), which increased the allocation in
cvm_oct_configure_common_hw()
from 128 to 1024, but left the corresponding teardown paths in cvm_oct_probe()
and cvm_oct_remove(), passing 128 to cvm_oct_mem_empty_fpa().

If anyone on the list with access to an Octeon board is able to test this, that
would be wonderful. Otherwise, I completely understand dropping this patch until
someone with hardware can verify it.

Thanks,
Priyanshu

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2026-11-16 10:16 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-16  6:57 [PATCH v3] staging: octeon: fix asymmetric output buffer pool accounting Priyanshu Singh
2026-11-16  9:40 ` Greg Kroah-Hartman
2026-11-16 10:16   ` ASH

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox