[PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[Thorsten Blum]

Use secs_to_jiffies() and simplify the code.

Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
---
 drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c
index 3dbd4efa21a2..19dce1ba512d 100644
--- a/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c
+++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c
@@ -220,7 +220,7 @@ static int hws_bwc_queue_poll(struct mlx5hws_context *ctx,
 			      bool drain)
 {
 	unsigned long timeout = jiffies +
-				msecs_to_jiffies(MLX5HWS_BWC_POLLING_TIMEOUT * MSEC_PER_SEC);
+				secs_to_jiffies(MLX5HWS_BWC_POLLING_TIMEOUT);
 	struct mlx5hws_flow_op_result comp[MLX5HWS_BWC_MATCHER_REHASH_BURST_TH];
 	u16 burst_th = hws_bwc_get_burst_th(ctx, queue_id);
 	bool got_comp = *pending_rules >= burst_th;
-- 
2.48.1

Re: [PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[Jacob Keller]

On 2/19/2025 12:49 PM, Thorsten Blum wrote:
> Use secs_to_jiffies() and simplify the code.
> 
> Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>

nit: this is a cleanup which should have the net-next prefix applied,
since this doesn't fix any user visible behavior.

Otherwise, seems like an ok change.

> ---
>  drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c b/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c
> index 3dbd4efa21a2..19dce1ba512d 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/steering/hws/bwc.c
> @@ -220,7 +220,7 @@ static int hws_bwc_queue_poll(struct mlx5hws_context *ctx,
>  			      bool drain)
>  {
>  	unsigned long timeout = jiffies +
> -				msecs_to_jiffies(MLX5HWS_BWC_POLLING_TIMEOUT * MSEC_PER_SEC);
> +				secs_to_jiffies(MLX5HWS_BWC_POLLING_TIMEOUT);
>  	struct mlx5hws_flow_op_result comp[MLX5HWS_BWC_MATCHER_REHASH_BURST_TH];

This looks like it violates RCT ordering now though.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>

Thanks,
Jake

>  	u16 burst_th = hws_bwc_get_burst_th(ctx, queue_id);
>  	bool got_comp = *pending_rules >= burst_th;

Re: [PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[Leon Romanovsky]

On Wed, Feb 19, 2025 at 03:45:02PM -0800, Jacob Keller wrote:
> 
> 
> On 2/19/2025 12:49 PM, Thorsten Blum wrote:
> > Use secs_to_jiffies() and simplify the code.
> > 
> > Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
> 
> nit: this is a cleanup which should have the net-next prefix applied,
> since this doesn't fix any user visible behavior.
> 
> Otherwise, seems like an ok change.

IMHO, completely useless change for old code. I can see a value in new
secs_to_jiffies() function for new code, but not for old code. I want
to believe that people who write kernel patches aware that 1000 msec
equal to 1 sec.

Thanks

Re: [PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[Thorsten Blum]

On 20. Feb 2025, at 08:13, Leon Romanovsky wrote:
> On Wed, Feb 19, 2025 at 03:45:02PM -0800, Jacob Keller wrote:
>> On 2/19/2025 12:49 PM, Thorsten Blum wrote:
>>> Use secs_to_jiffies() and simplify the code.
>>> 
>>> Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
>> 
>> nit: this is a cleanup which should have the net-next prefix applied,
>> since this doesn't fix any user visible behavior.
>> 
>> Otherwise, seems like an ok change.
> 
> IMHO, completely useless change for old code. I can see a value in new
> secs_to_jiffies() function for new code, but not for old code. I want
> to believe that people who write kernel patches aware that 1000 msec
> equal to 1 sec.

Using secs_to_jiffies() is shorter and requires less cognitive load to
read imo. Plus, it now fits within the preferred 80 columns limit.

This "old code" was added in d74ee6e197a2c ("net/mlx5: HWS, set timeout
on polling for completion") in January 2025.

Thanks,
Thorsten

Re: [PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[Leon Romanovsky]

On Thu, Feb 20, 2025 at 12:08:07PM +0100, Thorsten Blum wrote:
> On 20. Feb 2025, at 08:13, Leon Romanovsky wrote:
> > On Wed, Feb 19, 2025 at 03:45:02PM -0800, Jacob Keller wrote:
> >> On 2/19/2025 12:49 PM, Thorsten Blum wrote:
> >>> Use secs_to_jiffies() and simplify the code.
> >>> 
> >>> Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
> >> 
> >> nit: this is a cleanup which should have the net-next prefix applied,
> >> since this doesn't fix any user visible behavior.
> >> 
> >> Otherwise, seems like an ok change.
> > 
> > IMHO, completely useless change for old code. I can see a value in new
> > secs_to_jiffies() function for new code, but not for old code. I want
> > to believe that people who write kernel patches aware that 1000 msec
> > equal to 1 sec.
> 
> Using secs_to_jiffies() is shorter and requires less cognitive load to
> read imo. Plus, it now fits within the preferred 80 columns limit.

Unfortunately, I see this change as a churn and not an improvement.

> 
> This "old code" was added in d74ee6e197a2c ("net/mlx5: HWS, set timeout
> on polling for completion") in January 2025.

I got same conversion patches for RDMA.
https://lore.kernel.org/all/20250219-rdma-secs-to-jiffies-v1-0-b506746561a9@linux.microsoft.com

Thanks

> 
> Thanks,
> Thorsten
> 

Re: [PATCH] net/mlx5: Use secs_to_jiffies() instead of msecs_to_jiffies()

[Saeed Mahameed]

On 20 Feb 14:07, Leon Romanovsky wrote:
>On Thu, Feb 20, 2025 at 12:08:07PM +0100, Thorsten Blum wrote:
>> On 20. Feb 2025, at 08:13, Leon Romanovsky wrote:
>> > On Wed, Feb 19, 2025 at 03:45:02PM -0800, Jacob Keller wrote:
>> >> On 2/19/2025 12:49 PM, Thorsten Blum wrote:
>> >>> Use secs_to_jiffies() and simplify the code.
>> >>>
>> >>> Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
>> >>
>> >> nit: this is a cleanup which should have the net-next prefix applied,
>> >> since this doesn't fix any user visible behavior.
>> >>
>> >> Otherwise, seems like an ok change.
>> >
>> > IMHO, completely useless change for old code. I can see a value in new
>> > secs_to_jiffies() function for new code, but not for old code. I want
>> > to believe that people who write kernel patches aware that 1000 msec
>> > equal to 1 sec.
>>
>> Using secs_to_jiffies() is shorter and requires less cognitive load to
>> read imo. Plus, it now fits within the preferred 80 columns limit.

>Unfortunately, I see this change as a churn and not an improvement.

All patches with any justified improvement are welcome, as long as they are
not automated or bot generated spam.

This patch is reasonable, you need to add [PATCH net-next] prefix in
the title for this patch to get into the pipeline.

Feel free to add:
Reviewed-by: <saeed@kernel.org>

>
>>
>> This "old code" was added in d74ee6e197a2c ("net/mlx5: HWS, set timeout
>> on polling for completion") in January 2025.
>
>I got same conversion patches for RDMA.
>https://lore.kernel.org/all/20250219-rdma-secs-to-jiffies-v1-0-b506746561a9@linux.microsoft.com
>
>Thanks
>
>>
>> Thanks,
>> Thorsten
>>
>