powerpc: Fix ABIv2 issues with stack offsets in assembly code

Fix STK_PARAM and use it instead of hardcoding ABIv1 offsets.

Signed-off-by: Anton Blanchard <anton@samba.org>

powerpc: No need to use dot symbols when branching to a function

binutils is smart enough to know that a branch to a function
descriptor is actually a branch to the functions text addres

powerpc: No need to use dot symbols when branching to a function

binutils is smart enough to know that a branch to a function
descriptor is actually a branch to the functions text address.

Alan tells me that binutils has been doing this for 9 years.

Signed-off-by: Anton Blanchard <anton@samba.org>

show more ...

powerpc: Fix endian issues in VMX copy loops

Fix the permute loops for little endian.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <ben

powerpc: Fix endian issues in VMX copy loops

Fix the permute loops for little endian.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

show more ...

powerpc: Fix VMX fix for memcpy case

In 2fae7cdb60240e2e2d9b378afbf6d9fcce8a3890 ("powerpc: Fix VMX in
interrupt check in POWER7 copy loops"), Anton inadvertently
introduced a regres

powerpc: Fix VMX fix for memcpy case

In 2fae7cdb60240e2e2d9b378afbf6d9fcce8a3890 ("powerpc: Fix VMX in
interrupt check in POWER7 copy loops"), Anton inadvertently
introduced a regression for memcpy on POWER7 machines. copyuser and
memcpy diverge slightly in their use of cr1 (copyuser doesn't use it,
but memcpy does) and you end up clobbering that register with your fix.
That results in (taken from an FC18 kernel):

[ 18.824604] Unrecoverable VMX/Altivec Unavailable Exception f20 at c000000000052f40
[ 18.824618] Oops: Unrecoverable VMX/Altivec Unavailable Exception, sig: 6 [#1]
[ 18.824623] SMP NR_CPUS=1024 NUMA pSeries
[ 18.824633] Modules linked in: tg3(+) be2net(+) cxgb4(+) ipr(+) sunrpc xts lrw gf128mul dm_crypt dm_round_robin dm_multipath linear raid10 raid456 async_raid6_recov async_memcpy async_pq raid6_pq async_xor xor async_tx raid1 raid0 scsi_dh_rdac scsi_dh_hp_sw scsi_dh_emc scsi_dh_alua squashfs cramfs
[ 18.824705] NIP: c000000000052f40 LR: c00000000020b874 CTR: 0000000000000512
[ 18.824709] REGS: c000001f1fef7790 TRAP: 0f20 Not tainted (3.6.0-0.rc6.git0.2.fc18.ppc64)
[ 18.824713] MSR: 8000000000009032 <SF,EE,ME,IR,DR,RI> CR: 4802802e XER: 20000010
[ 18.824726] SOFTE: 0
[ 18.824728] CFAR: 0000000000000f20
[ 18.824731] TASK = c000000fa7128400[0] 'swapper/24' THREAD: c000000fa7480000 CPU: 24
GPR00: 00000000ffffffc0 c000001f1fef7a10 c00000000164edc0 c000000f9b9a8120
GPR04: c000000f9b9a8124 0000000000001438 0000000000000060 03ffffff064657ee
GPR08: 0000000080000000 0000000000000010 0000000000000020 0000000000000030
GPR12: 0000000028028022 c00000000ff25400 0000000000000001 0000000000000000
GPR16: 0000000000000000 7fffffffffffffff c0000000016b2180 c00000000156a500
GPR20: c000000f968c7a90 c0000000131c31d8 c000001f1fef4000 c000000001561d00
GPR24: 000000000000000a 0000000000000000 0000000000000001 0000000000000012
GPR28: c000000fa5c04f80 00000000000008bc c0000000015c0a28 000000000000022e
[ 18.824792] NIP [c000000000052f40] .memcpy_power7+0x5a0/0x7c4
[ 18.824797] LR [c00000000020b874] .pcpu_free_area+0x174/0x2d0
[ 18.824800] Call Trace:
[ 18.824803] [c000001f1fef7a10] [c000000000052c14] .memcpy_power7+0x274/0x7c4 (unreliable)
[ 18.824809] [c000001f1fef7b10] [c00000000020b874] .pcpu_free_area+0x174/0x2d0
[ 18.824813] [c000001f1fef7bb0] [c00000000020ba88] .free_percpu+0xb8/0x1b0
[ 18.824819] [c000001f1fef7c50] [c00000000043d144] .throtl_pd_exit+0x94/0xd0
[ 18.824824] [c000001f1fef7cf0] [c00000000043acf8] .blkg_free+0x88/0xe0
[ 18.824829] [c000001f1fef7d90] [c00000000018c048] .rcu_process_callbacks+0x2e8/0x8a0
[ 18.824835] [c000001f1fef7e90] [c0000000000a8ce8] .__do_softirq+0x158/0x4d0
[ 18.824840] [c000001f1fef7f90] [c000000000025ecc] .call_do_softirq+0x14/0x24
[ 18.824845] [c000000fa7483650] [c000000000010e80] .do_softirq+0x160/0x1a0
[ 18.824850] [c000000fa74836f0] [c0000000000a94a4] .irq_exit+0xf4/0x120
[ 18.824854] [c000000fa7483780] [c000000000020c44] .timer_interrupt+0x154/0x4d0
[ 18.824859] [c000000fa7483830] [c000000000003be0] decrementer_common+0x160/0x180
[ 18.824866] --- Exception: 901 at .plpar_hcall_norets+0x84/0xd4
[ 18.824866] LR = .check_and_cede_processor+0x48/0x80
[ 18.824871] [c000000fa7483b20] [c00000000007f018] .check_and_cede_processor+0x18/0x80 (unreliable)
[ 18.824877] [c000000fa7483b90] [c00000000007f104] .dedicated_cede_loop+0x84/0x150
[ 18.824883] [c000000fa7483c50] [c0000000006bc030] .cpuidle_enter+0x30/0x50
[ 18.824887] [c000000fa7483cc0] [c0000000006bc9f4] .cpuidle_idle_call+0x104/0x720
[ 18.824892] [c000000fa7483d80] [c000000000070af8] .pSeries_idle+0x18/0x40
[ 18.824897] [c000000fa7483df0] [c000000000019084] .cpu_idle+0x1a4/0x380
[ 18.824902] [c000000fa7483ec0] [c0000000008a4c18] .start_secondary+0x520/0x528
[ 18.824907] [c000000fa7483f90] [c0000000000093f0] .start_secondary_prolog+0x10/0x14
[ 18.824911] Instruction dump:
[ 18.824914] 38840008 90030000 90e30004 38630008 7ca62850 7cc300d0 78c7e102 7cf01120
[ 18.824923] 78c60660 39200010 39400020 39600030 <7e00200c> 7c0020ce 38840010 409f001c
[ 18.824935] ---[ end trace 0bb95124affaaa45 ]---
[ 18.825046] Unrecoverable VMX/Altivec Unavailable Exception f20 at c000000000052d08

I believe the right fix is to make memcpy match usercopy and not use
cr1.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: <stable@kernel.org> [v3.6]

show more ...

powerpc: Fix VMX in interrupt check in POWER7 copy loops

The enhanced prefetch hint patches corrupt the condition register
that was used to check if we are in interrupt. Fix this by usin

powerpc: Fix VMX in interrupt check in POWER7 copy loops

The enhanced prefetch hint patches corrupt the condition register
that was used to check if we are in interrupt. Fix this by using cr1.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

show more ...

powerpc: Merge STK_REG/PARAM/FRAMESIZE

Merge the defines of STACKFRAMESIZE, STK_REG, STK_PARAM from different
places.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Sign

powerpc: Merge STK_REG/PARAM/FRAMESIZE

Merge the defines of STACKFRAMESIZE, STK_REG, STK_PARAM from different
places.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

show more ...

powerpc: Fix usage of register macros getting ready for %r0 change

Anything that uses a constructed instruction (ie. from ppc-opcode.h),
need to use the new R0 macro, as %r0 is not going

powerpc: Fix usage of register macros getting ready for %r0 change

Anything that uses a constructed instruction (ie. from ppc-opcode.h),
need to use the new R0 macro, as %r0 is not going to work.

Also convert usages of macros where we are just determining an offset
(usually for a load/store), like:
std r14,STK_REG(r14)(r1)
Can't use STK_REG(r14) as %r14 doesn't work in the STK_REG macro since
it's just calculating an offset.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

show more ...

powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch

Implement a POWER7 optimised memcpy using VMX and enhanced prefetch
instructions.

This is a copy of the POWER7 o

powerpc: POWER7 optimised memcpy using VMX and enhanced prefetch

Implement a POWER7 optimised memcpy using VMX and enhanced prefetch
instructions.

This is a copy of the POWER7 optimised copy_to_user/copy_from_user
loop. Detailed implementation and performance details can be found in
commit a66086b8197d (powerpc: POWER7 optimised
copy_to_user/copy_from_user using VMX).

I noticed memcpy issues when profiling a RAID6 workload:

.memcpy
.async_memcpy
.async_copy_data
.__raid_run_ops
.handle_stripe
.raid5d
.md_thread

I created a simplified testcase by building a RAID6 array with 4 1GB
ramdisks (booting with brd.rd_size=1048576):

# mdadm -CR -e 1.2 /dev/md0 --level=6 -n4 /dev/ram[0-3]

I then timed how long it took to write to the entire array:

# dd if=/dev/zero of=/dev/md0 bs=1M

Before: 892 MB/s
After: 999 MB/s

A 12% improvement.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>

show more ...