Skip to contentSkip to conversation

igb: acknowledging time sync interrupts

DISCUSSION Richard Cochran 2 participants 4 messages May 26, 2013 – May 29, 2013

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

Matt,

In igb_main.c you have ISR code like:

	if (icr & E1000_ICR_TS) {
		u32 tsicr = rd32(E1000_TSICR);

		if (tsicr & E1000_TSICR_TXTS) {
			/* acknowledge the interrupt */
			wr32(E1000_TSICR, E1000_TSICR_TXTS);
			/* retrieve hardware timestamp */
			schedule_work(&adapter->ptp_tx_work);
		}
	}

In the datasheet for the 82580 and the i210, for TSICR it says,

Note: Once ICR.Time_Sync is set, the internal value of this
register should be cleared by writing 1b to all bits or cleared by a read to enable receiving an additional ^^^^^^^^^^^^^^^^^^^^
ICR.Time_Sync interrupt.

and that implies that your write to acknowledge the interrupt is superfluous, since you already read the TSICR.

Is this an error in the datasheets, or is the code doing extra, unneeded work?

Thanks,
Richard

#1 lore raw
Quoted 23 lines · Richard Cochran

Matt,

In igb_main.c you have ISR code like:

	if (icr & E1000_ICR_TS) {
		u32 tsicr = rd32(E1000_TSICR);

		if (tsicr & E1000_TSICR_TXTS) {
			/* acknowledge the interrupt */
			wr32(E1000_TSICR, E1000_TSICR_TXTS);
			/* retrieve hardware timestamp */
			schedule_work(&adapter->ptp_tx_work);
		}
	}

In the datasheet for the 82580 and the i210, for TSICR it says,

Note: Once ICR.Time_Sync is set, the internal value of this
register should be cleared by writing 1b to all bits or cleared by a read to enable receiving an additional ^^^^^^^^^^^^^^^^^^^^
ICR.Time_Sync interrupt.

Quoted 8 lines

and that implies that your write to acknowledge the interrupt is superfluous, since you already read the TSICR.

Is this an error in the datasheets, or is the code doing extra, unneeded work?

Thanks,
Richard

Richard,

Good catch--and you're correct--but I had issues with the read of TSICR not clearing like it should on the 82580 (but it would work fine on the I350 and I210). I decided the cleaner implementation would be to explicitly acknowledge the interrupt across the board. I haven't had the time to follow up with the hardware team, but my suspicion is that it's an errata with the 82580.

I'll start some internal discussion to see if I can get an official answer on the 82580.

Cheers,
Matthew

------------------------------------------------------------------------------
Try New Relic Now & We'll Send You this Cool Shirt
New Relic is the only SaaS-based application performance monitoring service that delivers powerful full stack analytics. Optimize and monitor your browser, app, & servers with just a few lines of code. Try New Relic
and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_may
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit http://communities.intel.com/community/wired

#2 lore raw
Quoted 3 lines · Matthew

I'll start some internal discussion to see if I can get an official answer on the 82580.

Okay, that sounds good. I just posted a patch that removes the wr32(), and I can report that it does work on the i210. If the 82580 needs this workaround, then I can add it back in again.

Thanks,
Richard

#3 lore raw
Quoted 6 lines · Richard Cochran
Quoted 4 lines · Matthew

I'll start some internal discussion to see if I can get an official answer
on the 82580.

Quoted 6 lines

Okay, that sounds good. I just posted a patch that removes the wr32(), and I can report that it does work on the i210. If the 82580 needs this workaround, then I can add it back in again.

Thanks,
Richard

Following up internally, we got confirmation 82580 does have an issue with the read of TSICR not clearing it. Please do add the wr32 back in and keep an eye out for updated documentation.

Cheers,
Matthew