Re: [PATCH net-next v3 2/2] bnxt_en: replace PTP spinlock with seqlock

[Vadim Fedorenko <vadim.fedorenko@linux.dev>]

On 29/10/2024 20:34, Michael Chan wrote:
> On Mon, Oct 28, 2024 at 11:57 AM Vadim Fedorenko <vadfed@meta.com> wrote:
>>
>> We can see high contention on ptp_lock while doing RX timestamping
>> on high packet rates over several queues. Spinlock is not effecient
>> to protect timecounter for RX timestamps when reads are the most
>> usual operations and writes are only occasional. It's better to use
>> seqlock in such cases.
>>
>> Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
>> ---
>> v3:
>> - remove unused variable
>> v2:
>> - use read_excl lock to serialize reg access with FW reset
>> ---
>>   drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 19 +++--
>>   drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c | 73 ++++++-------------
>>   drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.h | 14 +++-
>>   3 files changed, 46 insertions(+), 60 deletions(-)
>>
> 
>> diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c
>> index 820c7e83e586..5ab52f7a282d 100644
>> --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c
>> +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c
>> @@ -67,19 +67,21 @@ static int bnxt_ptp_settime(struct ptp_clock_info *ptp_info,
>>          if (BNXT_PTP_USE_RTC(ptp->bp))
>>                  return bnxt_ptp_cfg_settime(ptp->bp, ns);
>>
>> -       spin_lock_irqsave(&ptp->ptp_lock, flags);
>> +       write_seqlock_irqsave(&ptp->ptp_lock, flags);
>>          timecounter_init(&ptp->tc, &ptp->cc, ns);
>> -       spin_unlock_irqrestore(&ptp->ptp_lock, flags);
>> +       write_sequnlock_irqrestore(&ptp->ptp_lock, flags);
>>          return 0;
>>   }
>>
>> -/* Caller holds ptp_lock */
>>   static int bnxt_refclk_read(struct bnxt *bp, struct ptp_system_timestamp *sts,
>>                              u64 *ns)
>>   {
>>          struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
>>          u32 high_before, high_now, low;
>> +       unsigned long flags;
>>
>> +       /* We have to serialize reg access and FW reset */
>> +       read_seqlock_excl_irqsave(&ptp->ptp_lock, flags);
>>          if (test_bit(BNXT_STATE_IN_FW_RESET, &bp->state))
> 
> I think we need read_sequnlock_excl_irqrestore() here before returning.

Yes, you are right, we need to unlock seqlock in case of error. I'll
post v4 soon, thanks!

> 
>>                  return -EIO;
>>
>> @@ -93,6 +95,7 @@ static int bnxt_refclk_read(struct bnxt *bp, struct ptp_system_timestamp *sts,
>>                  low = readl(bp->bar0 + ptp->refclk_mapped_regs[0]);
>>                  ptp_read_system_postts(sts);
>>          }
>> +       read_sequnlock_excl_irqrestore(&ptp->ptp_lock, flags);
>>          *ns = ((u64)high_now << 32) | low;
>>
>>          return 0;