Search String: Display: Description: Sort:

Results:

References: [ +subject:/^(?:^\s*(re|sv|fwd|fw)[\[\]\d]*[:>-]+\s*)*\[openib\-general\]\s+Re\:\s+LLTX\s+and\s+netif_stop_queue\s*$/: 26 ]

Total 26 documents matching your query.

1. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: Grant Grundler <iod00d@xxxxxx>
Date: Mon, 3 Jan 2005 09:12:27 -0800
Cacheline bouncing usually starts to show up with > 4 CPUs (ie 8 or more). Some workloads that Jamal cares about (routing) only need 2 cpus. grant
/archives/netdev/2005-01/msg00043.html (11,395 bytes)

2. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: jamal <hadi@xxxxxxxxxx>
Date: 03 Jan 2005 23:18:14 -0500
What bothers me is more the complexity this has introduced more than the workload. OTOH, 1-2% improvement posted by Roland is good justification. cheers, jamal
/archives/netdev/2005-01/msg00055.html (11,638 bytes)

3. PATCH] kill needless addrconf lock in sctp. (score: 1)
Author: Olsson@xxxxxxxxxxx>
Date: Wed, 19 Jan 2005 14:47:11 -0800
ling from your tree at the moment though: Pull bk://kernel.bkbits.net/acme/connection_sock-2.6 -> file://disk1/BK/net-2.6 ERROR-Can'
/archives/netdev/2005-01/msg00862.html (11,869 bytes)

4. enib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: avem@xxxxxxxxxxxxx>
Date: Wed, 19 Jan 2005 15:18:53 -0800
t leave the SCTP code as-is for now.
/archives/netdev/2005-01/msg00863.html (12,903 bytes)

5. enib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: avem@xxxxxxxxxxxxx>
Date: Wed, 19 Jan 2005 15:41:32 -0800
X for non-loopback out of the tree.
/archives/netdev/2005-01/msg00864.html (13,534 bytes)

6. kge: new syskonnect gigabit ethernet driver (score: 1)
Author: ller" <davem@xxxxxxxxxxxxx>
Date: Wed, 19 Jan 2005 19:02:52 -0500
or that. Usually we strive for BH d
/archives/netdev/2005-01/msg00866.html (12,794 bytes)

7. kge: new syskonnect gigabit ethernet driver (score: 1)
Author: . Miller" <davem@xxxxxxxxxxxxx>
Date: Wed, 19 Jan 2005 16:46:14 -0800
rnel. But here, as soon as we call into ->hard_start_xmit(), the dr
/archives/netdev/2005-01/msg00868.html (12,993 bytes)

8. (score: 1)
Author: rzik@xxxxxxxxx>
Date: Wed, 19 Jan 2005 16:47:21 -0800
hen Hemminger <shemminger@xxxxxxxx>
/archives/netdev/2005-01/msg00870.html (12,549 bytes)

9. enib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: avamudan@xxxxxxxxx>
Date: Thu, 20 Jan 2005 01:47:53 +0100
thing like this might look like. In summary: 1) dev->xmit_lock is now IRQ disabling instead of BH disabling 2) Drivers can use dev->xmit_lock in place of their private driver_pri
/archives/netdev/2005-01/msg00871.html (12,489 bytes)

10. enib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: shemminger@xxxxxxxx>
Date: Wed, 19 Jan 2005 16:52:47 -0800
ype patch in a posting I just made.
/archives/netdev/2005-01/msg00872.html (13,288 bytes)

11. enib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: davem@xxxxxxxxxxxxx>
Date: Thu, 20 Jan 2005 02:17:30 +0100
for ordered ops on ring indexes so
/archives/netdev/2005-01/msg00873.html (12,117 bytes)

12. shared/cloned skbs (score: 1)
Author: dschuym@xxxxxxxxxx>
Date: Sat, 18 Dec 2004 09:58:15 -0800
by copying them before writing and dequeue unwriteable skbs unchanged. Assumes that IP/IPv6 header is always linear so no pulling required. Signed-off-by: Tho
/archives/netdev/2004-12/msg00496.html (9,980 bytes)

13. nd netif_stop_queue (score: 1)
Author: raf@xxxxxxx>
Date: Sat, 18 Dec 2004 10:26:40 -0800
just documenting that Roland> device drivers can use the xmit_lock member of struct Roland> net_device to synchronize other parts of the driver Roland> against
/archives/netdev/2004-12/msg00497.html (10,668 bytes)

14. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: Grant Grundler <iod00d@xxxxxx>
Date: Mon, 3 Jan 2005 09:12:27 -0800
Cacheline bouncing usually starts to show up with > 4 CPUs (ie 8 or more). Some workloads that Jamal cares about (routing) only need 2 cpus. grant
/archives/netdev/2005-01/msg01533.html (11,842 bytes)

15. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: jamal <hadi@xxxxxxxxxx>
Date: 03 Jan 2005 23:18:14 -0500
What bothers me is more the complexity this has introduced more than the workload. OTOH, 1-2% improvement posted by Roland is good justification. cheers, jamal
/archives/netdev/2005-01/msg01545.html (12,110 bytes)

16. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: "David S. Miller" <davem@xxxxxxxxxxxxx>
Date: Wed, 19 Jan 2005 14:47:11 -0800
I think I'm going to put in something like Eric's patch and fix up the other LLTX drivers as per his sungem patch. There is a part of me that does want to yank LLTX for non-loopback out of the tree.
/archives/netdev/2005-01/msg02352.html (12,376 bytes)

17. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: Stephen Hemminger <shemminger@xxxxxxxx>
Date: Wed, 19 Jan 2005 15:18:53 -0800
Wondering, why not just have the drivers have a way to lock dev->queue_lock in the interrupt handler, and change the xmit to do spin_lock_irqsave? Any driver that assumes it is being called with irq'
/archives/netdev/2005-01/msg02353.html (13,452 bytes)

18. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: "David S. Miller" <davem@xxxxxxxxxxxxx>
Date: Wed, 19 Jan 2005 15:41:32 -0800
Yes, that's an idea. Effectively, LLTX is working around the fact that dev->xmit_lock is BH disabling instead of IRQ disabling. And there is no fundamental reason for that. Usually we strive for BH d
/archives/netdev/2005-01/msg02354.html (14,108 bytes)

19. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: Jeff Garzik <jgarzik@xxxxxxxxx>
Date: Wed, 19 Jan 2005 19:02:52 -0500
David S. Miller wrote: Usually we strive for BH disabling locks in order to decrease the amount of hard IRQ disabling time in the kernel. But here, as soon as we call into ->hard_start_xmit(), the dr
/archives/netdev/2005-01/msg02356.html (13,439 bytes)

20. Re: [openib-general] Re: LLTX and netif_stop_queue (score: 1)
Author: Stephen Hemminger <shemminger@xxxxxxxx>
Date: Wed, 19 Jan 2005 16:46:14 -0800
But in the end it would be safer than the current LLTX usage in drivers which seems like it opening up problems. -- Stephen Hemminger <shemminger@xxxxxxxx>
/archives/netdev/2005-01/msg02358.html (13,621 bytes)


This search system is powered by Namazu