netdev - RE: [PATCH] ucc_geth: Ack IRQ events as late as possible.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 29 Apr 2008 20:14:14 +0800
From:	"Li Yang" <LeoLi@...escale.com>
To:	<joakim.tjernlund@...nsmode.se>
Cc:	"Netdev" <netdev@...r.kernel.org>
Subject: RE: [PATCH] ucc_geth: Ack IRQ events as late as possible.

> > > diff --git a/drivers/net/ucc_geth.h 
> b/drivers/net/ucc_geth.h index 
> > > cc24e87..7451f17 100644
> > > --- a/drivers/net/ucc_geth.h
> > > +++ b/drivers/net/ucc_geth.h
> > > @@ -207,8 +207,8 @@ struct ucc_geth {
> > >  			UCCE_RXB3 | UCCE_RXB2 | UCCE_RXB1 | UCCE_RXB0)
> > >  #define UCCE_RXF         (UCCE_RXF7 | UCCE_RXF6 | UCCE_RXF5 
> > > | UCCE_RXF4 |\
> > >  			UCCE_RXF3 | UCCE_RXF2 | UCCE_RXF1 | UCCE_RXF0)
> > > -#define UCCE_OTHER       (UCCE_SCAR | UCCE_GRA  | UCCE_CBPR 
> > > | UCCE_BSY  |\
> > > -			UCCE_RXC  | UCCE_TXC  | UCCE_TXE)
> > > +#define UCCE_OTHER       (UCCE_SCAR | UCCE_GRA | UCCE_CBPR |\
> > > +			UCCE_RXC  | UCCE_TXC)
> > 
> > This will remove UCCE_BSY and UCCE_TXE from the uccm_mask, hence 
> > disabled them from generating interrupts.
> 
> Yeah, got mislead by the UCCE_RX_EVENTS and UCC_TX_EVENTS just below.
> So the UCCE_OTHER change should go, but the rest stay.
> 
> > 
> > How is the improvement after your change to put off interrupt ack?
> > Isn't it caused by disabling the busy interrupts?
> 
> Not sure, got no time to measure ATM. Besides that, I do 

I agree that by putting off ack to TX/RX interrupts, there could be some
performance benefit in some cases.  But the last clear to ucce will make
it possible to miss a packet.

> think this change is needed anyway. There is no sane way to 
> handle the stats counters in the main irq loop. The next step 
> would be to actually start counting errors. Agreed?

I don't agree here.  The BSY event doesn't mean RXB will be set too.
The same applies for TX.  Acking the error events ealier will make the
error counter more accurate.

- Leo
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html