CONFIG_SMP: Status of pending patches?
Regarding two of Catalin's CONFIG_SMP patches in linux-arm git, I wonder what's the recent status regarding mainline inclusion? The two patches
1. Global ASID allocation on SMP
http://www.linux-arm.org/git?p=linux-2.6.git;a=commit;h=33618544e3151700950707c920ea2f7ec3e0b7ba
2. ARM: Implement read/write for ownership in the ARMv6 DMA cache ops
http://www.linux-arm.org/git?p=linux-2.6.git;a=commit;h=1bb211d4b81d65760d6befb17a01348db79459b1
seem to be essential for SMP stability.
Patch (1) is marked as applied [1]. Will it be part of final 2.6.33? For patch (2) I couldn't find anything. Is planned for mainline inclusion, too?
Sorry if I missed anything.
Thanks
Dirk
[1] http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=5905/1
Quoted 7 lines · Dirk Behme
Regarding two of Catalin's
CONFIG_SMPpatches in linux-arm git, I wonder what's the recent status regarding mainline inclusion? The two patches1. Global ASID allocation on SMP
http://www.linux-arm.org/git?p=linux-2.6.git;a=commit;h=33618544e3151700950707c920ea2f7ec3e0b7ba
Quoted 6 lines
2. ARM: Implement read/write for ownership in the ARMv6 DMA cache ops
http://www.linux-arm.org/git?p=linux-2.6.git;a=commit;h=1bb211d4b81d65760d6befb17a01348db79459b1seem to be essential for SMP stability.
Patch (1) is marked as applied [1]. Will it be part of final 2.6.33?
Patch 1 is queued for the next merge window.
Quoted 3 lines
For
patch (2) I couldn't find anything. Is planned for mainline inclusion, too?
I don't think patch 2 is up to being ready yet, and it's outdated with the respect to the rest of the DMA fixes.
Like it or not, 2.6.33 is not going to see a material improvement; the fixes for this are just too large and invasive for -rc kernels (as I pointed out before the previous merge window, when they ended up being dropped from that merge).
Unfortunately, they're still undergoing testing; TI are reporting weird failures on the patch set, but it's inconclusive at the moment - and it's looking like this testing might overrun this coming merge window. That means they could end up missing their second second merge window.
It annoys me that I can't test them myself in any reasonable and meaningful way - essentially my patch set is based on theory.
So, I'm going to continue sitting on them, not pushing them into mainline until we know what's going on, and whether any of this stuff is actually working properly.
They might get in for 2.6.35... I've just got no idea how long this is going to take to sort out.