[PATCH net-next 0/2] selftests: Couple of fixes in Toeplitz RPS cases

[PATCH net-next 0/2] selftests: Couple of fixes in Toeplitz RPS cases

[Gal Pressman]

Fix a couple of bugs in the RPS cases of the Toeplitz selftest.

Gal Pressman (2):
  selftests: drv-net: fix RPS mask handling in toeplitz test
  selftests: drv-net: fix RPS mask handling for high CPU numbers

 tools/testing/selftests/drivers/net/hw/toeplitz.c  | 4 ++--
 tools/testing/selftests/drivers/net/hw/toeplitz.py | 9 ++++++---
 2 files changed, 8 insertions(+), 5 deletions(-)

-- 
2.40.1

[PATCH net-next 1/2] selftests: drv-net: fix RPS mask handling in toeplitz test

[Gal Pressman]

The toeplitz.py test passed the hex mask without "0x" prefix (e.g.,
"300" for CPUs 8,9). The toeplitz.c strtoul() call wrongly parsed this
as decimal 300 (0x12c) instead of hex 0x300.

Use separate format strings for sysfs (plain hex via format()) and
command line (prefixed hex via hex()).

Fixes: 9cf9aa77a1f6 ("selftests: drv-net: hw: convert the Toeplitz test to Python")
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
---
 tools/testing/selftests/drivers/net/hw/toeplitz.py | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/drivers/net/hw/toeplitz.py b/tools/testing/selftests/drivers/net/hw/toeplitz.py
index d2db5ee9e358..7a9af4af1838 100755
--- a/tools/testing/selftests/drivers/net/hw/toeplitz.py
+++ b/tools/testing/selftests/drivers/net/hw/toeplitz.py
@@ -94,14 +94,17 @@ def _configure_rps(cfg, rps_cpus):
     mask = 0
     for cpu in rps_cpus:
         mask |= (1 << cpu)
-    mask = hex(mask)[2:]
+
+    # sysfs expect hex without '0x' prefix, toeplitz.c needs the prefix
+    mask_sysfs = format(mask, 'x')
+    mask_cmdline = hex(mask)
 
     # Set RPS bitmap for all rx queues
     for rps_file in glob.glob(f"/sys/class/net/{cfg.ifname}/queues/rx-*/rps_cpus"):
         with open(rps_file, "w", encoding="utf-8") as fp:
-            fp.write(mask)
+            fp.write(mask_sysfs)
 
-    return mask
+    return mask_cmdline
 
 
 def _send_traffic(cfg, proto_flag, ipver, port):
-- 
2.40.1

[PATCH net-next 2/2] selftests: drv-net: fix RPS mask handling for high CPU numbers

[Gal Pressman]

The RPS bitmask bounds check uses ~(RPS_MAX_CPUS - 1) which equals ~15 =
0xfff0, only allowing CPUs 0-3.

Change the mask to ~((1UL << RPS_MAX_CPUS) - 1) = ~0xffff to allow CPUs
0-15.

Fixes: 5ebfb4cc3048 ("selftests/net: toeplitz test")
Reviewed-by: Nimrod Oren <noren@nvidia.com>
Signed-off-by: Gal Pressman <gal@nvidia.com>
---
 tools/testing/selftests/drivers/net/hw/toeplitz.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/drivers/net/hw/toeplitz.c b/tools/testing/selftests/drivers/net/hw/toeplitz.c
index d23b3b0c20a3..285bb17df9c2 100644
--- a/tools/testing/selftests/drivers/net/hw/toeplitz.c
+++ b/tools/testing/selftests/drivers/net/hw/toeplitz.c
@@ -485,8 +485,8 @@ static void parse_rps_bitmap(const char *arg)
 
 	bitmap = strtoul(arg, NULL, 0);
 
-	if (bitmap & ~(RPS_MAX_CPUS - 1))
-		error(1, 0, "rps bitmap 0x%lx out of bounds 0..%lu",
+	if (bitmap & ~((1UL << RPS_MAX_CPUS) - 1))
+		error(1, 0, "rps bitmap 0x%lx out of bounds, max cpu %lu",
 		      bitmap, RPS_MAX_CPUS - 1);
 
 	for (i = 0; i < RPS_MAX_CPUS; i++)
-- 
2.40.1

Re: [PATCH net-next 2/2] selftests: drv-net: fix RPS mask handling for high CPU numbers

[Willem de Bruijn]

Gal Pressman wrote:
> The RPS bitmask bounds check uses ~(RPS_MAX_CPUS - 1) which equals ~15 =
> 0xfff0, only allowing CPUs 0-3.
> 
> Change the mask to ~((1UL << RPS_MAX_CPUS) - 1) = ~0xffff to allow CPUs
> 0-15.
> 
> Fixes: 5ebfb4cc3048 ("selftests/net: toeplitz test")
> Reviewed-by: Nimrod Oren <noren@nvidia.com>
> Signed-off-by: Gal Pressman <gal@nvidia.com>

Should go to net instead of net-next?

Reviewed-by: Willem de Bruijn <willemb@google.com> 

Re: [PATCH net-next 1/2] selftests: drv-net: fix RPS mask handling in toeplitz test

[Willem de Bruijn]

Gal Pressman wrote:
> The toeplitz.py test passed the hex mask without "0x" prefix (e.g.,
> "300" for CPUs 8,9). The toeplitz.c strtoul() call wrongly parsed this
> as decimal 300 (0x12c) instead of hex 0x300.
> 
> Use separate format strings for sysfs (plain hex via format()) and
> command line (prefixed hex via hex()).
> 
> Fixes: 9cf9aa77a1f6 ("selftests: drv-net: hw: convert the Toeplitz test to Python")
> Reviewed-by: Nimrod Oren <noren@nvidia.com>
> Signed-off-by: Gal Pressman <gal@nvidia.com>

Reviewed-by: Willem de Bruijn <willemb@google.com>

> ---
>  tools/testing/selftests/drivers/net/hw/toeplitz.py | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/drivers/net/hw/toeplitz.py b/tools/testing/selftests/drivers/net/hw/toeplitz.py
> index d2db5ee9e358..7a9af4af1838 100755
> --- a/tools/testing/selftests/drivers/net/hw/toeplitz.py
> +++ b/tools/testing/selftests/drivers/net/hw/toeplitz.py
> @@ -94,14 +94,17 @@ def _configure_rps(cfg, rps_cpus):
>      mask = 0
>      for cpu in rps_cpus:
>          mask |= (1 << cpu)
> -    mask = hex(mask)[2:]
> +
> +    # sysfs expect hex without '0x' prefix, toeplitz.c needs the prefix
> +    mask_sysfs = format(mask, 'x')

A particular reason not to use the existing slicing?

> +    mask_cmdline = hex(mask)
>  
>      # Set RPS bitmap for all rx queues
>      for rps_file in glob.glob(f"/sys/class/net/{cfg.ifname}/queues/rx-*/rps_cpus"):
>          with open(rps_file, "w", encoding="utf-8") as fp:
> -            fp.write(mask)
> +            fp.write(mask_sysfs)

Alternatively 

-    mask = hex(mask)[2:]
+    mask = hex(mask)

- fp.write(mask)
+ fp.write(mask[2:])

The comment that sysfs and toeplitz.c expect different input is
definitely helpful.

>  
> -    return mask
> +    return mask_cmdline
>  
>  
>  def _send_traffic(cfg, proto_flag, ipver, port):
> -- 
> 2.40.1
> 

Re: [PATCH net-next 2/2] selftests: drv-net: fix RPS mask handling for high CPU numbers

[Gal Pressman]

On 12/01/2026 5:29, Willem de Bruijn wrote:
> Gal Pressman wrote:
>> The RPS bitmask bounds check uses ~(RPS_MAX_CPUS - 1) which equals ~15 =
>> 0xfff0, only allowing CPUs 0-3.
>>
>> Change the mask to ~((1UL << RPS_MAX_CPUS) - 1) = ~0xffff to allow CPUs
>> 0-15.
>>
>> Fixes: 5ebfb4cc3048 ("selftests/net: toeplitz test")
>> Reviewed-by: Nimrod Oren <noren@nvidia.com>
>> Signed-off-by: Gal Pressman <gal@nvidia.com>
> 
> Should go to net instead of net-next?
> 
> Reviewed-by: Willem de Bruijn <willemb@google.com> 

I usually send tests bug fixes to net-next, since it doesn't fix a bug
in the kernel.

Should I send those to net instead?

Re: [PATCH net-next 1/2] selftests: drv-net: fix RPS mask handling in toeplitz test

[Gal Pressman]

On 12/01/2026 5:43, Willem de Bruijn wrote:
> Gal Pressman wrote:
>> The toeplitz.py test passed the hex mask without "0x" prefix (e.g.,
>> "300" for CPUs 8,9). The toeplitz.c strtoul() call wrongly parsed this
>> as decimal 300 (0x12c) instead of hex 0x300.
>>
>> Use separate format strings for sysfs (plain hex via format()) and
>> command line (prefixed hex via hex()).
>>
>> Fixes: 9cf9aa77a1f6 ("selftests: drv-net: hw: convert the Toeplitz test to Python")
>> Reviewed-by: Nimrod Oren <noren@nvidia.com>
>> Signed-off-by: Gal Pressman <gal@nvidia.com>
> 
> Reviewed-by: Willem de Bruijn <willemb@google.com>
> 
>> ---
>>  tools/testing/selftests/drivers/net/hw/toeplitz.py | 9 ++++++---
>>  1 file changed, 6 insertions(+), 3 deletions(-)
>>
>> diff --git a/tools/testing/selftests/drivers/net/hw/toeplitz.py b/tools/testing/selftests/drivers/net/hw/toeplitz.py
>> index d2db5ee9e358..7a9af4af1838 100755
>> --- a/tools/testing/selftests/drivers/net/hw/toeplitz.py
>> +++ b/tools/testing/selftests/drivers/net/hw/toeplitz.py
>> @@ -94,14 +94,17 @@ def _configure_rps(cfg, rps_cpus):
>>      mask = 0
>>      for cpu in rps_cpus:
>>          mask |= (1 << cpu)
>> -    mask = hex(mask)[2:]
>> +
>> +    # sysfs expect hex without '0x' prefix, toeplitz.c needs the prefix
>> +    mask_sysfs = format(mask, 'x')
> 
> A particular reason not to use the existing slicing?
> 
>> +    mask_cmdline = hex(mask)
>>  
>>      # Set RPS bitmap for all rx queues
>>      for rps_file in glob.glob(f"/sys/class/net/{cfg.ifname}/queues/rx-*/rps_cpus"):
>>          with open(rps_file, "w", encoding="utf-8") as fp:
>> -            fp.write(mask)
>> +            fp.write(mask_sysfs)
> 
> Alternatively 
> 
> -    mask = hex(mask)[2:]
> +    mask = hex(mask)
> 
> - fp.write(mask)
> + fp.write(mask[2:])
> 
> The comment that sysfs and toeplitz.c expect different input is
> definitely helpful.

Will take your suggestion, thanks.

Re: [PATCH net-next 2/2] selftests: drv-net: fix RPS mask handling for high CPU numbers

[Willem de Bruijn]

Gal Pressman wrote:
> On 12/01/2026 5:29, Willem de Bruijn wrote:
> > Gal Pressman wrote:
> >> The RPS bitmask bounds check uses ~(RPS_MAX_CPUS - 1) which equals ~15 =
> >> 0xfff0, only allowing CPUs 0-3.
> >>
> >> Change the mask to ~((1UL << RPS_MAX_CPUS) - 1) = ~0xffff to allow CPUs
> >> 0-15.
> >>
> >> Fixes: 5ebfb4cc3048 ("selftests/net: toeplitz test")
> >> Reviewed-by: Nimrod Oren <noren@nvidia.com>
> >> Signed-off-by: Gal Pressman <gal@nvidia.com>
> > 
> > Should go to net instead of net-next?
> > 
> > Reviewed-by: Willem de Bruijn <willemb@google.com> 
> 
> I usually send tests bug fixes to net-next, since it doesn't fix a bug
> in the kernel.
> 
> Should I send those to net instead?

I'm not aware of a separate policy for tests. Not sure if maintainers
have a preference. Probably behavior on this is a bit inconsistent
anyway.