NETIF_F_LLTX and race conditions [LWN.net]

NETIF_F_LLTX and race conditions

Posted Feb 3, 2005 15:40 UTC (Thu) by melauer (guest, #2438)
In reply to: NETIF_F_LLTX and race conditions by jwb
Parent article: NETIF_F_LLTX and race conditions

> With these network locking changes, and a brand new SCSI layer, I can
> hardly wait to roll out the new stable kernel on all my productions
> machines on the day it is released!

The definition of "stable kernel" has changed. The latest kernel release in an even-numbered series is not the "stable kernel" anymore. Now that releases which just fix bugs (e.g. security holes) and releases which add features have been thoroughly conflated, that's the way it's gotta be. The latest kernel release from your disto or hardware vendor is the "stable kernel" now. Presumably it's an older kernel with backported bugfixes.

(Log in to post comments)

stable kernel

Posted Feb 4, 2005 0:48 UTC (Fri) by giraffedata (subscriber, #1954) [Link]

Presumably it's an older kernel with backported bugfixes.

And that is, incidentally, probably based on 2.4.

stable kernel

Posted Feb 6, 2005 1:17 UTC (Sun) by barryn (subscriber, #5996) [Link]

> And that is, incidentally, probably based on 2.4.

Not if you're running any of the following distributions (and some others too):

Fedora Core 2 or 3
(once it comes out) Red Hat Enterprise Linux 4 (or recompiled clones thereof)
SuSE Linux Enterprise Server 9
SuSE Linux 9.1 or 9.2
Ubuntu
Mandrake 10.x

2.6 is slowly but steadily taking over...

stable kernel

Posted Feb 6, 2005 3:38 UTC (Sun) by giraffedata (subscriber, #1954) [Link]

I presume you're saying that the distributions mentioned are providing stabilized 2.6-based kernels, or recommending them, or abandonning support of 2.4-based kernels.

But I still maintain that if you find a stable Linux kernel, it's more likely to be based on 2.4, because these 2.6-based ones simply aren't stable in the way we got used to in the 2.4 days. The code in 2.6-based kernels is substantially newer and less exposed than in the 2.4-based ones.

I'm still hopeful that the distributions will stick with an old 2.6 level and let it stabilize, but so far I haven't seen the evidence that they will. If they frequently "upgrade" by grabbing all of Linus's recent changes, we'll still have to look to something 2.4-based for any kind of stability.