[RFC 00/20] Proposal for remaining BKL users
Mauro Carvalho Chehab
mchehab at redhat.com
Wed Jan 26 03:58:21 PST 2011
Em 26-01-2011 09:31, Arnd Bergmann escreveu:
> On Wednesday 26 January 2011, Greg KH wrote:
>> On Tue, Jan 25, 2011 at 11:17:14PM +0100, Arnd Bergmann wrote:
>>> I've gone through all the code in the kernel that
>>> uses the big kernel lock and come up with a solution
>>> that seems at least half-reasonable for each of them.
>>> The decisions are somewhat arbitrary, but here is
>>> what I'd suggest we do:
>>> * Remove in 2.6.39:
>>> i830, autofs3, smbfs
>> I thought some people really wanted to keep i830. Or was that i810?
>> I'll drop autofs3 and smbfs, thanks.
> i810 needs to be kept, i830 is obsolete, see the patch changelogs
> there. I assume that Dave Airlie will take both patches (i810 BKL
> removal and i830 removal) into his 2.6.39 queue.
>>> * Work around in an ugly way, but keep alive:
>>> * ufs, ipx, i810, cx25721
>>> * Fix properly:
>>> * usbip, go7007, adfs, x25
>> Thanks for the usbip and go7007 patches, I'll queue them up.
> It would be good if you could also take the cx25721 patch. It's
> not the nicest one I've done, but it's a bug fix nonetheless.
I guess you're meaning cx25821, right?
Palash should take a look on it and review. This is a device that allows
12 simultaneous streams, so, I suspect that he'll need to do some
changes at the locking schema, to avoid performance bottlenecks.
More information about the dri-devel