platform/kernel/linux-rpi.git
2 years agopowerpc/numa: Update cpu_cpu_map on CPU online/offline
Srikar Dronamraju [Thu, 26 Aug 2021 10:05:20 +0000 (15:35 +0530)]
powerpc/numa: Update cpu_cpu_map on CPU online/offline

cpu_cpu_map holds all the CPUs in the DIE. However in PowerPC, when
onlining/offlining of CPUs, this mask doesn't get updated.  This mask
is however updated when CPUs are added/removed. So when both
operations like online/offline of CPUs and adding/removing of CPUs are
done simultaneously, then cpumaps end up broken.

WARNING: CPU: 13 PID: 1142 at kernel/sched/topology.c:898
build_sched_domains+0xd48/0x1720
Modules linked in: rpadlpar_io rpaphp mptcp_diag xsk_diag tcp_diag
udp_diag raw_diag inet_diag unix_diag af_packet_diag netlink_diag
bonding tls nft_fib_inet nft_fib_ipv4 nft_fib_ipv6 nft_fib
nft_reject_inet nf_reject_ipv4 nf_reject_ipv6 nft_reject nft_ct
nft_chain_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 ip_set
rfkill nf_tables nfnetlink pseries_rng xts vmx_crypto uio_pdrv_genirq
uio binfmt_misc ip_tables xfs libcrc32c dm_service_time sd_mod t10_pi sg
ibmvfc scsi_transport_fc ibmveth dm_multipath dm_mirror dm_region_hash
dm_log dm_mod fuse
CPU: 13 PID: 1142 Comm: kworker/13:2 Not tainted 5.13.0-rc6+ #28
Workqueue: events cpuset_hotplug_workfn
NIP:  c0000000001caac8 LR: c0000000001caac4 CTR: 00000000007088ec
REGS: c00000005596f220 TRAP: 0700   Not tainted  (5.13.0-rc6+)
MSR:  8000000000029033 <SF,EE,ME,IR,DR,RI,LE>  CR: 48828222  XER:
00000009
CFAR: c0000000001ea698 IRQMASK: 0
GPR00: c0000000001caac4 c00000005596f4c0 c000000001c4a400 0000000000000036
GPR04: 00000000fffdffff c00000005596f1d0 0000000000000027 c0000018cfd07f90
GPR08: 0000000000000023 0000000000000001 0000000000000027 c0000018fe68ffe8
GPR12: 0000000000008000 c00000001e9d1880 c00000013a047200 0000000000000800
GPR16: c000000001d3c7d0 0000000000000240 0000000000000048 c000000010aacd18
GPR20: 0000000000000001 c000000010aacc18 c00000013a047c00 c000000139ec2400
GPR24: 0000000000000280 c000000139ec2520 c000000136c1b400 c000000001c93060
GPR28: c00000013a047c20 c000000001d3c6c0 c000000001c978a0 000000000000000d
NIP [c0000000001caac8] build_sched_domains+0xd48/0x1720
LR [c0000000001caac4] build_sched_domains+0xd44/0x1720
Call Trace:
[c00000005596f4c0] [c0000000001caac4] build_sched_domains+0xd44/0x1720 (unreliable)
[c00000005596f670] [c0000000001cc5ec] partition_sched_domains_locked+0x3ac/0x4b0
[c00000005596f710] [c0000000002804e4] rebuild_sched_domains_locked+0x404/0x9e0
[c00000005596f810] [c000000000283e60] rebuild_sched_domains+0x40/0x70
[c00000005596f840] [c000000000284124] cpuset_hotplug_workfn+0x294/0xf10
[c00000005596fc60] [c000000000175040] process_one_work+0x290/0x590
[c00000005596fd00] [c0000000001753c8] worker_thread+0x88/0x620
[c00000005596fda0] [c000000000181704] kthread+0x194/0x1a0
[c00000005596fe10] [c00000000000ccec] ret_from_kernel_thread+0x5c/0x70
Instruction dump:
485af049 60000000 2fa30800 409e0028 80fe0000 e89a00f8 e86100e8 38da0120
7f88e378 7ce53b78 4801fb91 60000000 <0fe0000039000000 38e00000 38c00000

Fix this by updating cpu_cpu_map aka cpumask_of_node() on every CPU
online/offline.

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100521.412639-5-srikar@linux.vnet.ibm.com
2 years agopowerpc/numa: Print debug statements only when required
Srikar Dronamraju [Thu, 26 Aug 2021 10:05:19 +0000 (15:35 +0530)]
powerpc/numa: Print debug statements only when required

Currently, a debug message gets printed every time an attempt to
add(remove) a CPU. However this is redundant if the CPU is already added
(removed) from the node.

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100521.412639-4-srikar@linux.vnet.ibm.com
2 years agopowerpc/numa: convert printk to pr_xxx
Srikar Dronamraju [Thu, 26 Aug 2021 10:05:18 +0000 (15:35 +0530)]
powerpc/numa: convert printk to pr_xxx

Convert the remaining printk to pr_xxx
One advantage would be all prints will now have prefix "numa:" from
pr_fmt().

[ convert printk(KERN_ERR) to pr_warn : Suggested by Laurent Dufour ]
Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
[mpe: Rebase onto powerpc/next, s/WARNING/Warning/]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100521.412639-3-srikar@linux.vnet.ibm.com
2 years agopowerpc/numa: Drop dbg in favour of pr_debug
Srikar Dronamraju [Thu, 26 Aug 2021 10:05:17 +0000 (15:35 +0530)]
powerpc/numa: Drop dbg in favour of pr_debug

powerpc supported numa=debug which is not documented. This option was
used to print early debug output. However something more flexible can be
achieved by using CONFIG_DYNAMIC_DEBUG.

Hence drop dbg (and numa=debug) in favour of pr_debug

Suggested-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
[mpe: Rebase on to powerpc/next form2 affinity changes]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100521.412639-2-srikar@linux.vnet.ibm.com
2 years agopowerpc/smp: Enable CACHE domain for shared processor
Srikar Dronamraju [Thu, 26 Aug 2021 10:04:01 +0000 (15:34 +0530)]
powerpc/smp: Enable CACHE domain for shared processor

Currently CACHE domain is not enabled on shared processor mode PowerVM
LPARS. On PowerVM systems, 'ibm,thread-group' device-tree property 2
under cpu-device-node indicates which all CPUs share L2-cache. However
'ibm,thread-group' device-tree property 2 is a relatively new property.
In absence of 'ibm,thread-group' property 2, 'l2-cache' device property
under cpu-device-node could help system to identify CPUs sharing L2-cache.
However this property is not exposed by PhyP in shared processor mode
configurations.

In absence of properties that inform OS about which CPUs share L2-cache,
fallback on core boundary.

Here are some stats from Power9 shared LPAR with the changes.

$ lscpu
Architecture:        ppc64le
Byte Order:          Little Endian
CPU(s):              32
On-line CPU(s) list: 0-31
Thread(s) per core:  8
Core(s) per socket:  1
Socket(s):           3
NUMA node(s):        2
Model:               2.2 (pvr 004e 0202)
Model name:          POWER9 (architected), altivec supported
Hypervisor vendor:   pHyp
Virtualization type: para
L1d cache:           32K
L1i cache:           32K
NUMA node0 CPU(s):   16-23
NUMA node1 CPU(s):   0-15,24-31
Physical sockets:    2
Physical chips:      1
Physical cores/chip: 10

Before patch
$ grep -r . /sys/kernel/debug/sched/domains/cpu0/domain*/name
Before
/sys/kernel/debug/sched/domains/cpu0/domain0/name:SMT
/sys/kernel/debug/sched/domains/cpu0/domain1/name:DIE
/sys/kernel/debug/sched/domains/cpu0/domain2/name:NUMA

After
/sys/kernel/debug/sched/domains/cpu0/domain0/name:SMT
/sys/kernel/debug/sched/domains/cpu0/domain1/name:CACHE
/sys/kernel/debug/sched/domains/cpu0/domain2/name:DIE
/sys/kernel/debug/sched/domains/cpu0/domain3/name:NUMA

$  awk '/domain/{print $1, $2}' /proc/schedstat | sort -u | sed -e 's/00000000,//g'
Before
domain0 00000055
domain0 000000aa
domain0 00005500
domain0 0000aa00
domain0 00550000
domain0 00aa0000
domain0 55000000
domain0 aa000000
domain1 00ff0000
domain1 ff00ffff
domain2 ffffffff

After
domain0 00000055
domain0 000000aa
domain0 00005500
domain0 0000aa00
domain0 00550000
domain0 00aa0000
domain0 55000000
domain0 aa000000
domain1 000000ff
domain1 0000ff00
domain1 00ff0000
domain1 ff000000
domain2 ff00ffff
domain2 ffffffff
domain3 ffffffff

(Lower is better)
perf stat -a -r 5 -n perf bench sched pipe  | tail -n 2
Before
           153.798 +- 0.142 seconds time elapsed  ( +-  0.09% )

After
           111.545 +- 0.652 seconds time elapsed  ( +-  0.58% )

which is an improvement of 27.47%

Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100401.412519-4-srikar@linux.vnet.ibm.com
2 years agopowerpc/smp: Update cpu_core_map on all PowerPc systems
Srikar Dronamraju [Thu, 26 Aug 2021 10:04:00 +0000 (15:34 +0530)]
powerpc/smp: Update cpu_core_map on all PowerPc systems

lscpu() uses core_siblings to list the number of sockets in the
system. core_siblings is set using topology_core_cpumask.

While optimizing the powerpc bootup path, Commit 4ca234a9cbd7
("powerpc/smp: Stop updating cpu_core_mask").  it was found that
updating cpu_core_mask() ended up taking a lot of time. It was thought
that on Powerpc, cpu_core_mask() would always be same as
cpu_cpu_mask() i.e number of sockets will always be equal to number of
nodes. As an optimization, cpu_core_mask() was made a snapshot of
cpu_cpu_mask().

However that was found to be false with PowerPc KVM guests, where each
node could have more than one socket. So with Commit c47f892d7aa6
("powerpc/smp: Reintroduce cpu_core_mask"), cpu_core_mask was updated
based on chip_id but in an optimized way using some mask manipulations
and chip_id caching.

However on non-PowerNV and non-pseries KVM guests (i.e not
implementing cpu_to_chip_id(), continued to use a copy of
cpu_cpu_mask().

There are two issues that were noticed on such systems
1. lscpu would report one extra socket.
On a IBM,9009-42A (aka zz system) which has only 2 chips/ sockets/
nodes, lscpu would report
Architecture:        ppc64le
Byte Order:          Little Endian
CPU(s):              160
On-line CPU(s) list: 0-159
Thread(s) per core:  8
Core(s) per socket:  6
Socket(s):           3                <--------------
NUMA node(s):        2
Model:               2.2 (pvr 004e 0202)
Model name:          POWER9 (architected), altivec supported
Hypervisor vendor:   pHyp
Virtualization type: para
L1d cache:           32K
L1i cache:           32K
L2 cache:            512K
L3 cache:            10240K
NUMA node0 CPU(s):   0-79
NUMA node1 CPU(s):   80-159

2. Currently cpu_cpu_mask is updated when a core is
added/removed. However its not updated when smt mode switching or on
CPUs are explicitly offlined. However all other percpu masks are
updated to ensure only active/online CPUs are in the masks.
This results in build_sched_domain traces since there will be CPUs in
cpu_cpu_mask() but those CPUs are not present in SMT / CACHE / MC /
NUMA domains. A loop of threads running smt mode switching and core
add/remove will soon show this trace.
Hence cpu_cpu_mask has to be update at smt mode switch.

This will have impact on cpu_core_mask(). cpu_core_mask() is a
snapshot of cpu_cpu_mask. Different CPUs within the same socket will
end up having different cpu_core_masks since they are snapshots at
different points of time. This means when lscpu will start reporting
many more sockets than the actual number of sockets/ nodes / chips.

Different ways to handle this problem:
A. Update the snapshot aka cpu_core_mask for all CPUs whenever
   cpu_cpu_mask is updated. This would a non-optimal solution.
B. Instead of a cpumask_var_t, make cpu_core_map a cpumask pointer
   pointing to cpu_cpu_mask. However percpu cpumask pointer is frowned
   upon and we need a clean way to handle PowerPc KVM guest which is
   not a snapshot.
C. Update cpu_core_masks all PowerPc systems like in PowerPc KVM
guests using mask manipulations. This approach is relatively simple
and unifies with the existing code.
D. On top of 3, we could also resurrect get_physical_package_id which
   could return a nid for the said CPU. However this is not needed at this
   time.

Option C is the preferred approach for now.

While this is somewhat a revert of Commit 4ca234a9cbd7 ("powerpc/smp:
Stop updating cpu_core_mask").

1. Plain revert has some conflicts
2. For chip_id == -1, the cpu_core_mask is made identical to
cpu_cpu_mask, unlike previously where cpu_core_mask was set to a core
if chip_id doesn't exist.

This goes by the principle that if chip_id is not exposed, then
sockets / chip / node share the same set of CPUs.

With the fix, lscpu o/p would be
Architecture:        ppc64le
Byte Order:          Little Endian
CPU(s):              160
On-line CPU(s) list: 0-159
Thread(s) per core:  8
Core(s) per socket:  6
Socket(s):           2                     <--------------
NUMA node(s):        2
Model:               2.2 (pvr 004e 0202)
Model name:          POWER9 (architected), altivec supported
Hypervisor vendor:   pHyp
Virtualization type: para
L1d cache:           32K
L1i cache:           32K
L2 cache:            512K
L3 cache:            10240K
NUMA node0 CPU(s):   0-79
NUMA node1 CPU(s):   80-159

Fixes: 4ca234a9cbd7 ("powerpc/smp: Stop updating cpu_core_mask")
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100401.412519-3-srikar@linux.vnet.ibm.com
2 years agopowerpc/smp: Fix a crash while booting kvm guest with nr_cpus=2
Srikar Dronamraju [Thu, 26 Aug 2021 10:03:59 +0000 (15:33 +0530)]
powerpc/smp: Fix a crash while booting kvm guest with nr_cpus=2

Aneesh reported a crash with a fairly recent upstream kernel when
booting kernel whose commandline was appended with nr_cpus=2

1:mon> e
cpu 0x1: Vector: 300 (Data Access) at [c000000008a67bd0]
    pc: c00000000002557c: cpu_to_chip_id+0x3c/0x100
    lr: c000000000058380: start_secondary+0x460/0xb00
    sp: c000000008a67e70
   msr: 8000000000001033
   dar: 10
 dsisr: 80000
  current = 0xc00000000891bb00
  paca    = 0xc0000018ff981f80   irqmask: 0x03   irq_happened: 0x01
    pid   = 0, comm = swapper/1
Linux version 5.13.0-rc3-15704-ga050a6d2b7e8 (kvaneesh@ltc-boston8) (gcc (Ubuntu 9.3.0-17ubuntu1~20.04) 9.3.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #433 SMP Tue May 25 02:38:49 CDT 2021
1:mon> t
[link register   ] c000000000058380 start_secondary+0x460/0xb00
[c000000008a67e70c000000008a67eb0 (unreliable)
[c000000008a67eb0c0000000000589d4 start_secondary+0xab4/0xb00
[c000000008a67f90c00000000000c654 start_secondary_prolog+0x10/0x14

Current code assumes that num_possible_cpus() is always greater than
threads_per_core. However this may not be true when using nr_cpus=2 or
similar options. Handle the case where num_possible_cpus() is not an
exact multiple of  threads_per_core.

Fixes: c1e53367dab1 ("powerpc/smp: Cache CPU to chip lookup")
Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Debugged-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826100401.412519-2-srikar@linux.vnet.ibm.com
2 years agopowerpc/configs/microwatt: Enable options for systemd
Joel Stanley [Thu, 26 Aug 2021 12:26:53 +0000 (21:56 +0930)]
powerpc/configs/microwatt: Enable options for systemd

When booting with systemd these options are required.

This increases the image by about 50KB, or 2%.

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826122653.3236867-4-joel@jms.id.au
2 years agopowerpc/configs/microwattt: Enable Liteeth
Joel Stanley [Thu, 26 Aug 2021 12:26:52 +0000 (21:56 +0930)]
powerpc/configs/microwattt: Enable Liteeth

Liteeth is the network device used by Microwatt.

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826122653.3236867-3-joel@jms.id.au
2 years agopowerpc/microwatt: Add Ethernet to device tree
Joel Stanley [Thu, 26 Aug 2021 12:26:51 +0000 (21:56 +0930)]
powerpc/microwatt: Add Ethernet to device tree

The liteeth network device is used in the Microwatt soc.

Signed-off-by: Joel Stanley <joel@jms.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210826122653.3236867-2-joel@jms.id.au
2 years agopowerpc: Redefine HMT_xxx macros as empty on PPC32
Christophe Leroy [Wed, 25 Aug 2021 13:34:45 +0000 (13:34 +0000)]
powerpc: Redefine HMT_xxx macros as empty on PPC32

HMT_xxx macros are macros for adjusting thread priority
(hardware multi-threading) are macros inherited from PPC64
via commit 5f7c690728ac ("[PATCH] powerpc: Merged ppc_asm.h")

Those instructions are pointless on PPC32, but some common
fonctions like arch_cpu_idle() use them.

So make them empty on PPC32 to avoid those instructions.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c5a07fadea33d640ad10cecf0ac8faaec1c524e0.1629898474.git.christophe.leroy@csgroup.eu
2 years agopowerpc/doc: Fix htmldocs errors
Aneesh Kumar K.V [Wed, 25 Aug 2021 04:24:47 +0000 (09:54 +0530)]
powerpc/doc: Fix htmldocs errors

Fix make htmldocs related errors with the newly added associativity.rst
doc file.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Tested-by: Stephen Rothwell <sfr@canb.auug.org.au> # build test
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210825042447.106219-1-aneesh.kumar@linux.ibm.com
2 years agoMerge changes from Paul Gortmaker
Michael Ellerman [Thu, 26 Aug 2021 14:49:06 +0000 (00:49 +1000)]
Merge changes from Paul Gortmaker

Merge the changes to retire the legacy WR sbc8548 and sbc8641 platforms
from Paul. These were sent as a pull request, but I rebased them onto
rc2 so as not to pull too many unrelated changes in to my next.

Description from Paul's pull request follows:

In v2.6.27 (2008, 917f0af9e5a9) the sbc8260 support was implicitly
retired by not being carried forward through the ppc --> powerpc
device tree transition.

Then, in v3.6 (2012, b048b4e17cbb) we retired the support for the
sbc8560 boards.

Next, in v4.18 (2017, 3bc6cf5a86e5) we retired the support for the
2006 vintage sbc834x boards.

The sbc8548 and sbc8641d boards were maybe 1-2 years newer than the
sbc834x boards, but it is also 3+ years later, so it makes sense to
now retire them as well - which is what is done here.

These two remaining WR boards were based on the Freescale MPC8548-CDS
and the MPC8641D-HPCN reference board implementations.  Having had the
chance to use these and many other Fsl ref boards, I know this:  The
Freescale reference boards were typically produced in limited quantity
and primarily available to BSP developers and hardware designers, and
not likely to have found a 2nd life with hobbyists and/or collectors.

It was good to have that BSP code subjected to mainline review and
hence also widely available back in the day. But given the above, we
should probably also be giving serious consideration to retiring
additional similar age/type reference board platforms as well.

I've always felt it is important for us to be proactive in retiring
old code, since it has a genuine non-zero carrying cost, as described
in the 930d52c012b8 merge log.  But for the here and now, we just
clean up the remaining BSP code that I had added for SBC platforms.

Link: https://lore.kernel.org/r/20210824174209.GB160508@windriver.com
2 years agoMAINTAINERS: update for Paul Gortmaker
Paul Gortmaker [Thu, 7 Jan 2021 19:18:59 +0000 (14:18 -0500)]
MAINTAINERS: update for Paul Gortmaker

Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2 years agopowerpc: retire sbc8641d board support
Paul Gortmaker [Thu, 7 Jan 2021 18:45:32 +0000 (13:45 -0500)]
powerpc: retire sbc8641d board support

The support was for this was added to mainline over 12 years ago, in
v2.6.26 [4e8aae89a35d] just around the ppc --> powerpc migration.

I believe the board was introduced shortly after the sbc8548 board,
making it roughly a 14 year old platform - with the CPU speed and
memory size typical for that era.

I haven't had one of these boards for several years, and availability
was discontinued several years before that.

Given that, there is no point in adding a burden to testing coverage
that builds all possible defconfigs, so it makes sense to remove it.

Of course it will remain in the git history forever, for anyone who
happens to find a functional board and wants to tinker with it.

Acked-by: Scott Wood <oss@buserror.net>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2 years agopowerpc: retire sbc8548 board support
Paul Gortmaker [Thu, 7 Jan 2021 13:40:38 +0000 (08:40 -0500)]
powerpc: retire sbc8548 board support

The support was for this was mainlined 13 years ago, in v2.6.25
[0e0fffe88767] just around the ppc --> powerpc migration.

I believe the board was introduced a year or two before that, so it
is roughly a 15 year old platform - with the CPU speed and memory size
that was typical for that era.

I haven't had one of these boards for several years, and availability
was discontinued several years before that.

Given that, there is no point in adding a burden to testing coverage
that builds all possible defconfigs, so it makes sense to remove it.

Of course it will remain in the git history forever, for anyone who
happens to find a functional board and wants to tinker with it.

Acked-by: Scott Wood <oss@buserror.net>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2 years agoMerge branch 'topic/ppc-kvm' into next
Michael Ellerman [Thu, 26 Aug 2021 11:21:11 +0000 (21:21 +1000)]
Merge branch 'topic/ppc-kvm' into next

Merge some KVM patches we are keeping in a topic branch in case there
are any merge conflicts that need resolving.

2 years agopowerpc: Refactor verification of MSR_RI
Christophe Leroy [Mon, 23 Aug 2021 08:24:21 +0000 (08:24 +0000)]
powerpc: Refactor verification of MSR_RI

40x and BOOKE don't have MSR_RI therefore all tests involving
MSR_RI may be problematic on those plateforms.

Create helpers to check or set MSR_RI in regs, and use them
in common code.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c2fb93708196734f4176dda334aaa3055f213b89.1629707037.git.christophe.leroy@csgroup.eu
2 years agopowerpc: Remove MSR_PR check in interrupt_exit_{user/kernel}_prepare()
Christophe Leroy [Mon, 23 Aug 2021 08:24:20 +0000 (08:24 +0000)]
powerpc: Remove MSR_PR check in interrupt_exit_{user/kernel}_prepare()

In those hot functions that are called at every interrupt, any saved
cycle is worth it.

interrupt_exit_user_prepare() and interrupt_exit_kernel_prepare() are
called from three places:
- From entry_32.S
- From interrupt_64.S
- From interrupt_exit_user_restart() and interrupt_exit_kernel_restart()

In entry_32.S, there are inambiguously called based on MSR_PR:

interrupt_return:
lwz r4,_MSR(r1)
addi r3,r1,STACK_FRAME_OVERHEAD
andi. r0,r4,MSR_PR
beq .Lkernel_interrupt_return
bl interrupt_exit_user_prepare
...
.Lkernel_interrupt_return:
bl interrupt_exit_kernel_prepare

In interrupt_64.S, that's similar:

interrupt_return_\srr\():
ld r4,_MSR(r1)
andi. r0,r4,MSR_PR
beq interrupt_return_\srr\()_kernel
interrupt_return_\srr\()_user: /* make backtraces match the _kernel variant */
addi r3,r1,STACK_FRAME_OVERHEAD
bl interrupt_exit_user_prepare
...
interrupt_return_\srr\()_kernel:
addi r3,r1,STACK_FRAME_OVERHEAD
bl interrupt_exit_kernel_prepare

In interrupt_exit_user_restart() and interrupt_exit_kernel_restart(),
MSR_PR is verified respectively by BUG_ON(!user_mode(regs)) and
BUG_ON(user_mode(regs)) prior to calling interrupt_exit_user_prepare()
and interrupt_exit_kernel_prepare().

The verification in interrupt_exit_user_prepare() and
interrupt_exit_kernel_prepare() are therefore useless and can be removed.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Acked-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/385ead49ccb66a259b25fee3eebf0bd4094068f3.1629707037.git.christophe.leroy@csgroup.eu
2 years agopowerpc/64e: Get dear offset with _DEAR macro
Xiongwei Song [Sat, 7 Aug 2021 01:02:39 +0000 (09:02 +0800)]
powerpc/64e: Get dear offset with _DEAR macro

Use _DEAR to get the offset of dear register in pr_regs for 64e cpus.

Signed-off-by: Xiongwei Song <sxwjean@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210807010239.416055-5-sxwjean@me.com
2 years agopowerpc: Add dear as a synonym for pt_regs.dar register
Xiongwei Song [Sat, 7 Aug 2021 01:02:38 +0000 (09:02 +0800)]
powerpc: Add dear as a synonym for pt_regs.dar register

Create an anonymous union for dar and dear regsiters, we can reference
dear to get the effective address when CONFIG_4xx=y or CONFIG_BOOKE=y.
Otherwise, reference dar. This makes code more clear.

Signed-off-by: Xiongwei Song <sxwjean@gmail.com>
[mpe: Reword commit title]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210807010239.416055-4-sxwjean@me.com
2 years agopowerpc/64e: Get esr offset with _ESR macro
Xiongwei Song [Sat, 7 Aug 2021 01:02:37 +0000 (09:02 +0800)]
powerpc/64e: Get esr offset with _ESR macro

Use _ESR to get the offset of esr register in pr_regs for 64e cpus.

Signed-off-by: Xiongwei Song <sxwjean@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210807010239.416055-3-sxwjean@me.com
2 years agopowerpc: Add esr as a synonym for pt_regs.dsisr
Xiongwei Song [Sat, 7 Aug 2021 01:02:36 +0000 (09:02 +0800)]
powerpc: Add esr as a synonym for pt_regs.dsisr

Create an anonymous union for dsisr and esr regsiters, we can reference
esr to get the exception detail when CONFIG_4xx=y or CONFIG_BOOKE=y.
Otherwise, reference dsisr. This makes code more clear.

Signed-off-by: Xiongwei Song <sxwjean@gmail.com>
[mpe: Reword commit title]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210807010239.416055-2-sxwjean@me.com
2 years agoselftests: Skip TM tests on synthetic TM implementations
Jordan Niethe [Thu, 29 Jul 2021 04:13:17 +0000 (14:13 +1000)]
selftests: Skip TM tests on synthetic TM implementations

Transactional Memory was removed from the architecture in ISA v3.1. For
threads running in P8/P9 compatibility mode on P10 a synthetic TM
implementation is provided. In this implementation, tbegin. always sets
cr0 eq meaning the abort handler is always called. This is not an issue
as users of TM are expected to have a fallback non transactional way to
make forward progress in the abort handler.  The TEXASR indicates if a
transaction failure is due to a synthetic implementation.

Some of the TM self tests need a non-degenerate TM implementation for
their testing to be meaningful so check for a synthetic implementation
and skip the test if so.

Signed-off-by: Jordan Niethe <jniethe5@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210729041317.366612-2-jniethe5@gmail.com
2 years agoselftests/powerpc: Add missing clobbered register to to ptrace TM tests
Jordan Niethe [Thu, 29 Jul 2021 04:13:16 +0000 (14:13 +1000)]
selftests/powerpc: Add missing clobbered register to to ptrace TM tests

ISA v3.1 removes TM but includes a synthetic implementation for
backwards compatibility.  With this implementation,  the tests
ptrace-tm-spd-gpr and ptrace-tm-gpr should never be able to make any
forward progress and eventually should be killed by the timeout.
Instead on a P10 running in P9 mode, ptrace_tm_gpr fails like so:

test: ptrace_tm_gpr
tags: git_version:unknown
Starting the child
...
...
GPR[27]: 1 Expected: 2
GPR[28]: 1 Expected: 2
GPR[29]: 1 Expected: 2
GPR[30]: 1 Expected: 2
GPR[31]: 1 Expected: 2
[FAIL] Test FAILED on line 98
failure: ptrace_tm_gpr
selftests:  ptrace-tm-gpr [FAIL]

The problem is in the inline assembly of the child. r0 is loaded with a
value in the child's transaction abort handler but this register is not
included in the clobbers list.  This means it is possible that this
statement:
cptr[1] = 0;
which is meant to signal the parent to wait may actually use the value
placed into r0 by the inline assembly incorrectly signal the parent to
continue.

By inspection the same problem is present in ptrace-tm-spd-gpr.

Adding r0 to the clobbbers list makes the test fail correctly via a
timeout on a P10 running in P8/P9 compatibility mode.

Suggested-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Jordan Niethe <jniethe5@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210729041317.366612-1-jniethe5@gmail.com
2 years agopowerpc/perf: Fix the check for SIAR value
Kajol Jain [Wed, 18 Aug 2021 17:15:56 +0000 (22:45 +0530)]
powerpc/perf: Fix the check for SIAR value

Incase of random sampling, there can be scenarios where
Sample Instruction Address Register(SIAR) may not latch
to the sampled instruction and could result in
the value of 0. In these scenarios it is preferred to
return regs->nip. These corner cases are seen in the
previous generation (p9) also.

Patch adds the check for SIAR value along with regs_use_siar
and siar_valid checks so that the function will return
regs->nip incase SIAR is zero.

Patch drops the code under PPMU_P10_DD1 flag check
which handles SIAR 0 case only for Power10 DD1.

Fixes: 2ca13a4cc56c9 ("powerpc/perf: Use regs->nip when SIAR is zero")
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210818171556.36912-3-kjain@linux.ibm.com
2 years agopowerpc/perf: Drop the case of returning 0 as instruction pointer
Kajol Jain [Wed, 18 Aug 2021 17:15:55 +0000 (22:45 +0530)]
powerpc/perf: Drop the case of returning 0 as instruction pointer

Drop the case of returning 0 as instruction pointer since kernel
never executes at 0 and userspace almost never does either.

Fixes: e6878835ac47 ("powerpc/perf: Sample only if SIAR-Valid bit is set in P7+")
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210818171556.36912-2-kjain@linux.ibm.com
2 years agopowerpc/perf: Use stack siar instead of mfspr
Kajol Jain [Wed, 18 Aug 2021 17:15:54 +0000 (22:45 +0530)]
powerpc/perf: Use stack siar instead of mfspr

Minor optimization in the 'perf_instruction_pointer' function code by
making use of stack siar instead of mfspr.

Fixes: 75382aa72f06 ("powerpc/perf: Move code to select SIAR or pt_regs into perf_read_regs")
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Tested-by: Nageswara R Sastry <rnsastry@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210818171556.36912-1-kjain@linux.ibm.com
2 years agopowerpc/64s: Remove WORT SPR from POWER9/10
Nicholas Piggin [Wed, 11 Aug 2021 16:00:44 +0000 (02:00 +1000)]
powerpc/64s: Remove WORT SPR from POWER9/10

This register is not architected and not implemented in POWER9 or 10,
it just reads back zeroes for compatibility.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Fabiano Rosas <farosas@linux.ibm.com>
Link: https://lore.kernel.org/r/20210811160134.904987-11-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV Nested: Reflect guest PMU in-use to L0 when guest SPRs are live
Nicholas Piggin [Wed, 11 Aug 2021 16:00:43 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV Nested: Reflect guest PMU in-use to L0 when guest SPRs are live

After the L1 saves its PMU SPRs but before loading the L2's PMU SPRs,
switch the pmcregs_in_use field in the L1 lppaca to the value advertised
by the L2 in its VPA. On the way out of the L2, set it back after saving
the L2 PMU registers (if they were in-use).

This transfers the PMU liveness indication between the L1 and L2 at the
points where the registers are not live.

This fixes the nested HV bug for which a workaround was added to the L0
HV by commit 63279eeb7f93a ("KVM: PPC: Book3S HV: Always save guest pmu
for guest capable of nesting"), which explains the problem in detail.
That workaround is no longer required for guests that include this bug
fix.

Fixes: 360cae313702 ("KVM: PPC: Book3S HV: Nested guest entry via hypercall")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Fabiano Rosas <farosas@linux.ibm.com>
Link: https://lore.kernel.org/r/20210811160134.904987-10-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV Nested: save_hv_return_state does not require trap argument
Fabiano Rosas [Wed, 11 Aug 2021 16:00:42 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV Nested: save_hv_return_state does not require trap argument

vcpu is already anargument so vcpu->arch.trap can be used directly.

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-9-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV Nested: Stop forwarding all HFUs to L1
Fabiano Rosas [Wed, 11 Aug 2021 16:00:41 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV Nested: Stop forwarding all HFUs to L1

If the nested hypervisor has no access to a facility because it has
been disabled by the host, it should also not be able to see the
Hypervisor Facility Unavailable that arises from one of its guests
trying to access the facility.

This patch turns a HFU that happened in L2 into a Hypervisor Emulation
Assistance interrupt and forwards it to L1 for handling. The ones that
happened because L1 explicitly disabled the facility for L2 are still
let through, along with the corresponding Cause bits in the HFSCR.

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
[np: move handling into kvmppc_handle_nested_exit]
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-8-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV Nested: Make nested HFSCR state accessible
Nicholas Piggin [Wed, 11 Aug 2021 16:00:40 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV Nested: Make nested HFSCR state accessible

When the L0 runs a nested L2, there are several permutations of HFSCR
that can be relevant. The HFSCR that the L1 vcpu L1 requested, the
HFSCR that the L1 vcpu may use, and the HFSCR that is actually being
used to run the L2.

The L1 requested HFSCR is not accessible outside the nested hcall
handler, so copy that into a new kvm_nested_guest.hfscr field.

The permitted HFSCR is taken from the HFSCR that the L1 runs with,
which is also not accessible while the hcall is being made. Move
this into a new kvm_vcpu_arch.hfscr_permitted field.

These will be used by the next patch to improve facility handling
for nested guests, and later by facility demand faulting patches.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-7-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV Nested: Sanitise vcpu registers
Fabiano Rosas [Wed, 11 Aug 2021 16:00:39 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV Nested: Sanitise vcpu registers

As one of the arguments of the H_ENTER_NESTED hypercall, the nested
hypervisor (L1) prepares a structure containing the values of various
hypervisor-privileged registers with which it wants the nested guest
(L2) to run. Since the nested HV runs in supervisor mode it needs the
host to write to these registers.

To stop a nested HV manipulating this mechanism and using a nested
guest as a proxy to access a facility that has been made unavailable
to it, we have a routine that sanitises the values of the HV registers
before copying them into the nested guest's vcpu struct.

However, when coming out of the guest the values are copied as they
were back into L1 memory, which means that any sanitisation we did
during guest entry will be exposed to L1 after H_ENTER_NESTED returns.

This patch alters this sanitisation to have effect on the vcpu->arch
registers directly before entering and after exiting the guest,
leaving the structure that is copied back into L1 unchanged (except
when we really want L1 to access the value, e.g the Cause bits of
HFSCR).

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Link: https://lore.kernel.org/r/20210811160134.904987-6-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV Nested: Fix TM softpatch HFAC interrupt emulation
Nicholas Piggin [Wed, 11 Aug 2021 16:00:38 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV Nested: Fix TM softpatch HFAC interrupt emulation

Have the TM softpatch emulation code set up the HFAC interrupt and
return -1 in case an instruction was executed with HFSCR bits clear,
and have the interrupt exit handler fall through to the HFAC handler.
When the L0 is running a nested guest, this ensures the HFAC interrupt
is correctly passed up to the L1.

The "direct guest" exit handler will turn these into PROGILL program
interrupts so functionality in practice will be unchanged. But it's
possible an L1 would want to handle these in a different way.

Also rearrange the FAC interrupt emulation code to match the HFAC format
while here (mainly, adding the FSCR_INTR_CAUSE mask).

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-5-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV P9: Fixes for TM softpatch interrupt NIP
Nicholas Piggin [Wed, 11 Aug 2021 16:00:37 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV P9: Fixes for TM softpatch interrupt NIP

The softpatch interrupt sets HSRR0 to the faulting instruction +4, so
it should subtract 4 for the faulting instruction address in the case
it is a TM softpatch interrupt (the instruction was not executed) and
it was not emulated.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-4-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV: Remove TM emulation from POWER7/8 path
Nicholas Piggin [Wed, 11 Aug 2021 16:00:36 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV: Remove TM emulation from POWER7/8 path

TM fake-suspend emulation is only used by POWER9. Remove it from the old
code path.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-3-npiggin@gmail.com
2 years agoKVM: PPC: Book3S HV: Initialise vcpu MSR with MSR_ME
Nicholas Piggin [Wed, 11 Aug 2021 16:00:35 +0000 (02:00 +1000)]
KVM: PPC: Book3S HV: Initialise vcpu MSR with MSR_ME

It is possible to create a VCPU without setting the MSR before running
it, which results in a warning in kvmhv_vcpu_entry_p9() that MSR_ME is
not set. This is pretty harmless because the MSR_ME bit is added to
HSRR1 before HRFID to guest, and a normal qemu guest doesn't hit it.

Initialise the vcpu MSR with MSR_ME set.

Reported-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210811160134.904987-2-npiggin@gmail.com
2 years agopowerpc: Use lwarx/ldarx directly instead of PPC_LWARX/LDARX macros
Christophe Leroy [Tue, 2 Mar 2021 08:48:11 +0000 (08:48 +0000)]
powerpc: Use lwarx/ldarx directly instead of PPC_LWARX/LDARX macros

Force the eh flag at 0 on PPC32.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/1fc81f07cabebb875b963e295408cc3dd38c8d85.1614674882.git.christophe.leroy@csgroup.eu
2 years agopowerpc/ptrace: Make user_mode() common to PPC32 and PPC64
Christophe Leroy [Tue, 17 Aug 2021 16:00:14 +0000 (16:00 +0000)]
powerpc/ptrace: Make user_mode() common to PPC32 and PPC64

Today we have:

#ifdef __powerpc64__
#define user_mode(regs) ((((regs)->msr) >> MSR_PR_LG) & 0x1)
#else
#define user_mode(regs) (((regs)->msr & MSR_PR) != 0)
#endif

With ppc64_defconfig, we get:

if (!user_mode(regs))
    14b4: e9 3e 01 08  ld      r9,264(r30)
    14b8: 71 29 40 00  andi.   r9,r9,16384
    14bc: 41 82 07 a4  beq     1c60 <.emulate_instruction+0x7d0>

If taking the ppc32 definition of user_mode(), the exact same code
is generated for ppc64_defconfig.

So, only keep one version of user_mode(), preferably the one not
using MSR_PR_LG which should be kept internal to reg.h.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/000a28c51808bbd802b505af42d2cb316c2be7d3.1629216000.git.christophe.leroy@csgroup.eu
2 years agopowerpc/syscalls: Simplify do_mmap2()
Christophe Leroy [Fri, 25 Jun 2021 10:58:33 +0000 (10:58 +0000)]
powerpc/syscalls: Simplify do_mmap2()

When shift is nul, operations remain valid so no test needed.

And 'ret' is unnecessary.

And use IS_ALIGNED() to check alignment, that's more clear.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/373ec500f386374bc5735007df3d3869eac47be1.1624618701.git.christophe.leroy@csgroup.eu
2 years agopowerpc/ptdump: Convert powerpc to GENERIC_PTDUMP
Christophe Leroy [Thu, 8 Jul 2021 16:49:43 +0000 (16:49 +0000)]
powerpc/ptdump: Convert powerpc to GENERIC_PTDUMP

This patch converts powerpc to the generic PTDUMP implementation.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/03166d569526be70214fe9370a7bad219d2f41c8.1625762907.git.christophe.leroy@csgroup.eu
2 years agopowerpc/ptdump: Reduce level numbers by 1 in note_page() and add p4d level
Christophe Leroy [Thu, 8 Jul 2021 16:49:42 +0000 (16:49 +0000)]
powerpc/ptdump: Reduce level numbers by 1 in note_page() and add p4d level

Do the same as commit f8f0d0b6fa20 ("mm: ptdump: reduce level numbers
by 1 in note_page()") and add missing p4d level.

This will align powerpc to the users of generic ptdump.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/d76495c574132b197b445a1f133755cca4b912a4.1625762906.git.christophe.leroy@csgroup.eu
2 years agopowerpc/ptdump: Remove unused 'page_size' parameter
Christophe Leroy [Thu, 8 Jul 2021 16:49:41 +0000 (16:49 +0000)]
powerpc/ptdump: Remove unused 'page_size' parameter

note_page_update_state() doesn't use page_size. Remove it.

Could also be removed to note_page() but as a following patch
will remove all current users of note_page(), just leave it as
is for now.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/e2f80d052001155251bfe009c360d0c5d9242c6b.1625762906.git.christophe.leroy@csgroup.eu
2 years agopowerpc/ptdump: Use DEFINE_SHOW_ATTRIBUTE()
Christophe Leroy [Thu, 8 Jul 2021 16:49:40 +0000 (16:49 +0000)]
powerpc/ptdump: Use DEFINE_SHOW_ATTRIBUTE()

Use DEFINE_SHOW_ATTRIBUTE() instead of open coding
open() and fops.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/b864a92693ca8413ef0b19f0c12065c212899b6e.1625762905.git.christophe.leroy@csgroup.eu
2 years agopowerpc: Avoid link stack corruption in misc asm functions
Christophe Leroy [Tue, 24 Aug 2021 07:56:35 +0000 (07:56 +0000)]
powerpc: Avoid link stack corruption in misc asm functions

bl;mflr is used at several places to get code position.

Use bcl 20,31,+4 instead of bl in order to preserve link stack.

See commit c974809a26a1 ("powerpc/vdso: Avoid link stack corruption
in __get_datapage()") for details.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c6eabb4fb6c156f75d56dcbcc6f243e5ac0fba42.1629791763.git.christophe.leroy@csgroup.eu
2 years agopowerpc/booke: Avoid link stack corruption in several places
Christophe Leroy [Tue, 24 Aug 2021 07:56:26 +0000 (07:56 +0000)]
powerpc/booke: Avoid link stack corruption in several places

Use bcl 20,31,+4 instead of bl in order to preserve link stack.

See commit c974809a26a1 ("powerpc/vdso: Avoid link stack corruption
in __get_datapage()") for details.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/e9fbc285eceb720e6c0e032ef47fe8b05f669b48.1629791751.git.christophe.leroy@csgroup.eu
2 years agopowerpc/32: indirect function call use bctrl rather than blrl in ret_from_kernel_thread
Christophe Leroy [Fri, 20 Aug 2021 05:16:05 +0000 (05:16 +0000)]
powerpc/32: indirect function call use bctrl rather than blrl in ret_from_kernel_thread

Copied from commit 89bbe4c798bc ("powerpc/64: indirect function call
use bctrl rather than blrl in ret_from_kernel_thread")

blrl is not recommended to use as an indirect function call, as it may
corrupt the link stack predictor.

This is not a performance critical path but this should be fixed for
consistency.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/91b1d242525307ceceec7ef6e832bfbacdd4501b.1629436472.git.christophe.leroy@csgroup.eu
2 years agoKVM: PPC: Book3S HV: Stop exporting symbols from book3s_64_mmu_radix
Fabiano Rosas [Thu, 5 Aug 2021 21:26:16 +0000 (18:26 -0300)]
KVM: PPC: Book3S HV: Stop exporting symbols from book3s_64_mmu_radix

The book3s_64_mmu_radix.o object is not part of the KVM builtins and
all the callers of the exported symbols are in the same kvm-hv.ko
module so we should not need to export any symbols.

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210805212616.2641017-4-farosas@linux.ibm.com
2 years agoKVM: PPC: Book3S HV: Add sanity check to copy_tofrom_guest
Fabiano Rosas [Thu, 5 Aug 2021 21:26:15 +0000 (18:26 -0300)]
KVM: PPC: Book3S HV: Add sanity check to copy_tofrom_guest

Both paths into __kvmhv_copy_tofrom_guest_radix ensure that we arrive
with an effective address that is smaller than our total addressable
space and addresses quadrant 0.

- The H_COPY_TOFROM_GUEST hypercall path rejects the call with
H_PARAMETER if the effective address has any of the twelve most
significant bits set.

- The kvmhv_copy_tofrom_guest_radix path clears the top twelve bits
before calling the internal function.

Although the callers make sure that the effective address is sane, any
future use of the function is exposed to a programming error, so add a
sanity check.

Suggested-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210805212616.2641017-3-farosas@linux.ibm.com
2 years agoKVM: PPC: Book3S HV: Fix copy_tofrom_guest routines
Fabiano Rosas [Thu, 5 Aug 2021 21:26:14 +0000 (18:26 -0300)]
KVM: PPC: Book3S HV: Fix copy_tofrom_guest routines

The __kvmhv_copy_tofrom_guest_radix function was introduced along with
nested HV guest support. It uses the platform's Radix MMU quadrants to
provide a nested hypervisor with fast access to its nested guests
memory (H_COPY_TOFROM_GUEST hypercall). It has also since been added
as a fast path for the kvmppc_ld/st routines which are used during
instruction emulation.

The commit def0bfdbd603 ("powerpc: use probe_user_read() and
probe_user_write()") changed the low level copy function from
raw_copy_from_user to probe_user_read, which adds a check to
access_ok. In powerpc that is:

 static inline bool __access_ok(unsigned long addr, unsigned long size)
 {
         return addr < TASK_SIZE_MAX && size <= TASK_SIZE_MAX - addr;
 }

and TASK_SIZE_MAX is 0x0010000000000000UL for 64-bit, which means that
setting the two MSBs of the effective address (which correspond to the
quadrant) now cause access_ok to reject the access.

This was not caught earlier because the most common code path via
kvmppc_ld/st contains a fallback (kvm_read_guest) that is likely to
succeed for L1 guests. For nested guests there is no fallback.

Another issue is that probe_user_read (now __copy_from_user_nofault)
does not return the number of bytes not copied in case of failure, so
the destination memory is not being cleared anymore in
kvmhv_copy_from_guest_radix:

 ret = kvmhv_copy_tofrom_guest_radix(vcpu, eaddr, to, NULL, n);
 if (ret > 0)                            <-- always false!
         memset(to + (n - ret), 0, ret);

This patch fixes both issues by skipping access_ok and open-coding the
low level __copy_to/from_user_inatomic.

Fixes: def0bfdbd603 ("powerpc: use probe_user_read() and probe_user_write()")
Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210805212616.2641017-2-farosas@linux.ibm.com
2 years agopowerpc/compat_sys: Declare syscalls
Cédric Le Goater [Mon, 23 Aug 2021 09:00:39 +0000 (11:00 +0200)]
powerpc/compat_sys: Declare syscalls

This fixes a compile error with W=1.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210823090039.166120-3-clg@kaod.org
2 years agopowerpc/prom: Fix unused variable ‘reserve_map’ when CONFIG_PPC32 is not set
Cédric Le Goater [Mon, 23 Aug 2021 09:00:38 +0000 (11:00 +0200)]
powerpc/prom: Fix unused variable ‘reserve_map’ when CONFIG_PPC32 is not set

This fixes a compile error with W=1.

arch/powerpc/kernel/prom.c: In function ‘early_reserve_mem’:
arch/powerpc/kernel/prom.c:625:10: error: variable ‘reserve_map’ set but not used [-Werror=unused-but-set-variable]
  __be64 *reserve_map;
          ^~~~~~~~~~~
cc1: all warnings being treated as errors

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210823090039.166120-2-clg@kaod.org
2 years agopowerpc/syscalls: Remove __NR__exit
Christophe Leroy [Mon, 23 Aug 2021 06:45:20 +0000 (06:45 +0000)]
powerpc/syscalls: Remove __NR__exit

__NR__exit is nowhere used. On most architectures it was removed by
commit 135ab6ec8fda ("[PATCH] remove remaining errno and
__KERNEL_SYSCALLS__ references") but not on powerpc.

powerpc removed __KERNEL_SYSCALLS__ in commit 3db03b4afb3e ("[PATCH]
rename the provided execve functions to kernel_execve"), but __NR__exit
was left over.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/6457eb4f327313323ed1f70e540bbb4ddc9178fa.1629701106.git.christophe.leroy@csgroup.eu
2 years agopowerpc/audit: Simplify syscall_get_arch()
Christophe Leroy [Fri, 20 Aug 2021 09:39:14 +0000 (09:39 +0000)]
powerpc/audit: Simplify syscall_get_arch()

Make use of is_32bit_task() and CONFIG_CPU_LITTLE_ENDIAN
to simplify syscall_get_arch().

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/4be53b9187a4d8c163968f4d224267e41a7fcc33.1629451479.git.christophe.leroy@csgroup.eu
2 years agopowerpc/audit: Avoid unneccessary #ifdef in syscall_get_arguments()
Christophe Leroy [Fri, 20 Aug 2021 09:28:19 +0000 (09:28 +0000)]
powerpc/audit: Avoid unneccessary #ifdef in syscall_get_arguments()

Use is_32bit_task() which already handles CONFIG_COMPAT.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/ba49cdd574558a0363300c3f6b5b062b397cb071.1629451483.git.christophe.leroy@csgroup.eu
2 years agoKVM: PPC: Book3S PR: Remove unused variable
Cédric Le Goater [Thu, 19 Aug 2021 12:56:54 +0000 (14:56 +0200)]
KVM: PPC: Book3S PR: Remove unused variable

This fixes a compile error with W=1.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210819125656.14498-5-clg@kaod.org
2 years agoKVM: PPC: Book3S PR: Declare kvmppc_handle_exit_pr()
Cédric Le Goater [Thu, 19 Aug 2021 12:56:53 +0000 (14:56 +0200)]
KVM: PPC: Book3S PR: Declare kvmppc_handle_exit_pr()

This fixes a compile error with W=1.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210819125656.14498-4-clg@kaod.org
2 years agopowerpc/pseries/vas: Declare pseries_vas_fault_thread_fn() as static
Cédric Le Goater [Thu, 19 Aug 2021 12:56:52 +0000 (14:56 +0200)]
powerpc/pseries/vas: Declare pseries_vas_fault_thread_fn() as static

This fixes a compile error with W=1.

Fixes: 6d0aaf5e0de0 ("powerpc/pseries/vas: Setup IRQ and fault handling")
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210819125656.14498-3-clg@kaod.org
2 years agopowerpc/perf/hv-gpci: Fix counter value parsing
Kajol Jain [Fri, 13 Aug 2021 08:21:58 +0000 (13:51 +0530)]
powerpc/perf/hv-gpci: Fix counter value parsing

H_GetPerformanceCounterInfo (0xF080) hcall returns the counter data in
the result buffer. Result buffer has specific format defined in the PAPR
specification. One of the fields is counter offset and width of the
counter data returned.

Counter data are returned in a unsigned char array in big endian byte
order. To get the final counter data, the values must be left shifted
byte at a time. But commit 220a0c609ad17 ("powerpc/perf: Add support for
the hv gpci (get performance counter info) interface") made the shifting
bitwise and also assumed little endian order. Because of that, hcall
counters values are reported incorrectly.

In particular this can lead to counters go backwards which messes up the
counter prev vs now calculation and leads to huge counter value
reporting:

  #: perf stat -e hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
           -C 0 -I 1000
        time             counts unit events
     1.000078854 18,446,744,073,709,535,232      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     2.000213293                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     3.000320107                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     4.000428392                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     5.000537864                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     6.000649087                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     7.000760312                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     8.000865218             16,448      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
     9.000978985 18,446,744,073,709,535,232      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
    10.001088891             16,384      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
    11.001201435                  0      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/
    12.001307937 18,446,744,073,709,535,232      hv_gpci/system_tlbie_count_and_time_tlbie_instructions_issued/

Fix the shifting logic to correct match the format, ie. read bytes in
big endian order.

Fixes: e4f226b1580b ("powerpc/perf/hv-gpci: Increase request buffer size")
Cc: stable@vger.kernel.org # v4.6+
Reported-by: Nageswara R Sastry<rnsastry@linux.ibm.com>
Signed-off-by: Kajol Jain <kjain@linux.ibm.com>
Tested-by: Nageswara R Sastry<rnsastry@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210813082158.429023-1-kjain@linux.ibm.com
2 years agopowerpc/tau: Add 'static' storage qualifier to 'tau_work' definition
Finn Thain [Thu, 19 Aug 2021 00:46:54 +0000 (10:46 +1000)]
powerpc/tau: Add 'static' storage qualifier to 'tau_work' definition

This patch prevents the following sparse warning.

arch/powerpc/kernel/tau_6xx.c:199:1: sparse: sparse: symbol 'tau_work'
was not declared. Should it be static?

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/44ab381741916a51e783c4a50d0b186abdd8f280.1629334014.git.fthain@linux-m68k.org
2 years agopowerpc/kvm: Remove obsolete and unneeded select
Lukas Bulwahn [Thu, 19 Aug 2021 11:39:53 +0000 (13:39 +0200)]
powerpc/kvm: Remove obsolete and unneeded select

Commit a278e7ea608b ("powerpc: Fix compile issue with force DAWR")
selects the non-existing config PPC_DAWR_FORCE_ENABLE for config
KVM_BOOK3S_64_HANDLER. As this commit also introduces a config PPC_DAWR
and this config PPC_DAWR is selected with PPC if PPC64, there is no
need for any further select in the KVM_BOOK3S_64_HANDLER.

Remove an obsolete and unneeded select in config KVM_BOOK3S_64_HANDLER.

The issue was identified with ./scripts/checkkconfigsymbols.py.

Fixes: a278e7ea608b ("powerpc: Fix compile issue with force DAWR")
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Reviewed-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210819113954.17515-2-lukas.bulwahn@gmail.com
2 years agopowerpc/32: Remove unneccessary calculations in load_up_{fpu/altivec}
Christophe Leroy [Wed, 18 Aug 2021 08:47:28 +0000 (08:47 +0000)]
powerpc/32: Remove unneccessary calculations in load_up_{fpu/altivec}

No need to re-read SPRN_THREAD, we can calculate thread address
from current (r2).

And remove a reload of value 1 into r4 as r4 is already 1.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/c04cce578b97a76a9e69a096698b1d89f721768a.1629276437.git.christophe.leroy@csgroup.eu
2 years agoselftests/powerpc: Remove duplicated include from tm-poison.c
Zheng Yongjun [Fri, 26 Mar 2021 06:48:08 +0000 (14:48 +0800)]
selftests/powerpc: Remove duplicated include from tm-poison.c

Remove duplicated include.

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zheng Yongjun <zhengyongjun3@huawei.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210326064808.3262568-1-zhengyongjun3@huawei.com
2 years agopowerpc: Remove duplicate includes
Wan Jiabing [Tue, 23 Mar 2021 06:29:05 +0000 (14:29 +0800)]
powerpc: Remove duplicate includes

interrupt.c: asm/interrupt.h has been included at line 12, so remove the
duplicate one at line 10.

time.c: linux/sched/clock.h has been included at line 33,so remove the
duplicate one at line 56 and move sched/cputime.h under sched including
segament.

Signed-off-by: Wan Jiabing <wanjiabing@vivo.com>
Reviewed-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210323062916.295346-1-wanjiabing@vivo.com
2 years agopowerpc/configs: Regenerate mpc885_ads_defconfig
Joel Stanley [Tue, 17 Aug 2021 04:54:07 +0000 (14:24 +0930)]
powerpc/configs: Regenerate mpc885_ads_defconfig

Regenerate atop v5.14-rc6 by doing a make savedefconfig.

The changes a re-ordering except for the following (which are still set
indirectly):

 - CONFIG_DEBUG_KERNEL=y selected by EXPERT

 - CONFIG_PPC_EARLY_DEBUG_CPM_ADDR=0xff002008 which is the default
   setting

Signed-off-by: Joel Stanley <joel@jms.id.au>
Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210817045407.2445664-4-joel@jms.id.au
2 years agopowerpc/config: Renable MTD_PHYSMAP_OF
Joel Stanley [Tue, 17 Aug 2021 04:54:06 +0000 (14:24 +0930)]
powerpc/config: Renable MTD_PHYSMAP_OF

CONFIG_MTD_PHYSMAP_OF is not longer enabled as it depends on
MTD_PHYSMAP which is not enabled.

This is a regression from commit 642b1e8dbed7 ("mtd: maps: Merge
physmap_of.c into physmap-core.c"), which added the extra dependency.
Add CONFIG_MTD_PHYSMAP=y so this stays in the config, as Christophe said
it is useful for build coverage.

Fixes: 642b1e8dbed7 ("mtd: maps: Merge physmap_of.c into physmap-core.c")
Signed-off-by: Joel Stanley <joel@jms.id.au>
Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210817045407.2445664-3-joel@jms.id.au
2 years agopowerpc/config: Fix IPV6 warning in mpc855_ads
Joel Stanley [Tue, 17 Aug 2021 04:54:05 +0000 (14:24 +0930)]
powerpc/config: Fix IPV6 warning in mpc855_ads

When building this config there's a warning:

  79:warning: override: reassigning to symbol IPV6

Commit 9a1762a4a4ff ("powerpc/8xx: Update mpc885_ads_defconfig to
improve CI") added CONFIG_IPV6=y, but left '# CONFIG_IPV6 is not set'
in.

IPV6 is default y, so remove both to clean up the build.

Fixes: 9a1762a4a4ff ("powerpc/8xx: Update mpc885_ads_defconfig to improve CI")
Signed-off-by: Joel Stanley <joel@jms.id.au>
Acked-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210817045407.2445664-2-joel@jms.id.au
2 years agopowerpc/head_check: Fix shellcheck errors
Michael Ellerman [Mon, 16 Aug 2021 06:36:02 +0000 (16:36 +1000)]
powerpc/head_check: Fix shellcheck errors

Replace "cat file | grep pattern" with "grep pattern file", and quote a
few variables. Together that fixes all shellcheck errors.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210817125154.3369884-1-mpe@ellerman.id.au
2 years agopowerpc/head_check: use stdout for error messages
Randy Dunlap [Sun, 15 Aug 2021 22:23:34 +0000 (15:23 -0700)]
powerpc/head_check: use stdout for error messages

Prefer stderr instead of stdout for error messages.
This is a good practice and can help CI error detecting and
reporting (0day in this case).

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210815222334.9575-1-rdunlap@infradead.org
2 years agopowerpc/pseries: Fix build error when NUMA=n
Michael Ellerman [Mon, 16 Aug 2021 02:30:11 +0000 (12:30 +1000)]
powerpc/pseries: Fix build error when NUMA=n

As reported by lkp, if NUMA=n we see a build error:

   arch/powerpc/platforms/pseries/hotplug-cpu.c: In function 'pseries_cpu_hotplug_init':
   arch/powerpc/platforms/pseries/hotplug-cpu.c:1022:8: error: 'node_to_cpumask_map' undeclared
    1022 |        node_to_cpumask_map[node]);

Use cpumask_of_node() which has an empty stub for NUMA=n, and when
NUMA=y does a lookup from node_to_cpumask_map[].

Fixes: bd1dd4c5f528 ("powerpc/pseries: Prevent free CPU ids being reused on another node")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210816041032.2839343-1-mpe@ellerman.id.au
2 years agopowerpc: Add "-z notext" flag to disable diagnostic
Fangrui Song [Fri, 13 Aug 2021 20:05:11 +0000 (13:05 -0700)]
powerpc: Add "-z notext" flag to disable diagnostic

Object files used to link .tmp_vmlinux.kallsyms1 have many
R_PPC64_ADDR64 relocations in non-SHF_WRITE sections. There are many
text relocations (e.g. in .rela___ksymtab_gpl+* and .rela__mcount_loc
sections) in a -pie link and are disallowed by LLD:

  ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output
  >>> defined in arch/powerpc/kernel/head_64.o
  >>> referenced by arch/powerpc/kernel/head_64.o:(__restart_table+0x10)

Newer GNU ld configured with "--enable-textrel-check=error" will report
an error as well:

  $ ld-new -EL -m elf64lppc -pie ... -o .tmp_vmlinux.kallsyms1 ...
  ld-new: read-only segment has dynamic relocations

Add "-z notext" to suppress the errors. Non-CONFIG_RELOCATABLE builds
use the default -no-pie mode and thus R_PPC64_ADDR64 relocations can be
resolved at link-time.

Reported-by: Itaru Kitayama <itaru.kitayama@riken.jp>
Co-developed-by: Bill Wendling <morbo@google.com>
Signed-off-by: Fangrui Song <maskray@google.com>
Signed-off-by: Bill Wendling <morbo@google.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210813200511.1905703-1-morbo@google.com
2 years agopowerpc/bug: Provide better flexibility to WARN_ON/__WARN_FLAGS() with asm goto
Christophe Leroy [Tue, 13 Apr 2021 16:38:10 +0000 (16:38 +0000)]
powerpc/bug: Provide better flexibility to WARN_ON/__WARN_FLAGS() with asm goto

Using asm goto in __WARN_FLAGS() and WARN_ON() allows more
flexibility to GCC.

For that add an entry to the exception table so that
program_check_exception() knowns where to resume execution
after a WARNING.

Here are two exemples. The first one is done on PPC32 (which
benefits from the previous patch), the second is on PPC64.

unsigned long test(struct pt_regs *regs)
{
int ret;

WARN_ON(regs->msr & MSR_PR);

return regs->gpr[3];
}

unsigned long test9w(unsigned long a, unsigned long b)
{
if (WARN_ON(!b))
return 0;
return a / b;
}

Before the patch:

000003a8 <test>:
 3a8: 81 23 00 84  lwz     r9,132(r3)
 3ac: 71 29 40 00  andi.   r9,r9,16384
 3b0: 40 82 00 0c  bne     3bc <test+0x14>
 3b4: 80 63 00 0c  lwz     r3,12(r3)
 3b8: 4e 80 00 20  blr

 3bc: 0f e0 00 00  twui    r0,0
 3c0: 80 63 00 0c  lwz     r3,12(r3)
 3c4: 4e 80 00 20  blr

0000000000000bf0 <.test9w>:
 bf0: 7c 89 00 74  cntlzd  r9,r4
 bf4: 79 29 d1 82  rldicl  r9,r9,58,6
 bf8: 0b 09 00 00  tdnei   r9,0
 bfc: 2c 24 00 00  cmpdi   r4,0
 c00: 41 82 00 0c  beq     c0c <.test9w+0x1c>
 c04: 7c 63 23 92  divdu   r3,r3,r4
 c08: 4e 80 00 20  blr

 c0c: 38 60 00 00  li      r3,0
 c10: 4e 80 00 20  blr

After the patch:

000003a8 <test>:
 3a8: 81 23 00 84  lwz     r9,132(r3)
 3ac: 71 29 40 00  andi.   r9,r9,16384
 3b0: 40 82 00 0c  bne     3bc <test+0x14>
 3b4: 80 63 00 0c  lwz     r3,12(r3)
 3b8: 4e 80 00 20  blr

 3bc: 0f e0 00 00  twui    r0,0

0000000000000c50 <.test9w>:
 c50: 7c 89 00 74  cntlzd  r9,r4
 c54: 79 29 d1 82  rldicl  r9,r9,58,6
 c58: 0b 09 00 00  tdnei   r9,0
 c5c: 7c 63 23 92  divdu   r3,r3,r4
 c60: 4e 80 00 20  blr

 c70: 38 60 00 00  li      r3,0
 c74: 4e 80 00 20  blr

In the first exemple, we see GCC doesn't need to duplicate what
happens after the trap.

In the second exemple, we see that GCC doesn't need to emit a test
and a branch in the likely path in addition to the trap.

We've got some WARN_ON() in .softirqentry.text section so it needs
to be added in the OTHER_TEXT_SECTIONS in modpost.c

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/389962b1b702e3c78d169e59bcfac56282889173.1618331882.git.christophe.leroy@csgroup.eu
2 years agopowerpc/bug: Remove specific powerpc BUG_ON() and WARN_ON() on PPC32
Christophe Leroy [Tue, 13 Apr 2021 16:38:09 +0000 (16:38 +0000)]
powerpc/bug: Remove specific powerpc BUG_ON() and WARN_ON() on PPC32

powerpc BUG_ON() and WARN_ON() are based on using twnei instruction.

For catching simple conditions like a variable having value 0, this
is efficient because it does the test and the trap at the same time.
But most conditions used with BUG_ON or WARN_ON are more complex and
forces GCC to format the condition into a 0 or 1 value in a register.
This will usually require 2 to 3 instructions.

The most efficient solution would be to use __builtin_trap() because
GCC is able to optimise the use of the different trap instructions
based on the requested condition, but this is complex if not
impossible for the following reasons:
- __builtin_trap() is a non-recoverable instruction, so it can't be
used for WARN_ON
- Knowing which line of code generated the trap would require the
analysis of DWARF information. This is not a feature we have today.

As mentioned in commit 8d4fbcfbe0a4 ("Fix WARN_ON() on bitfield ops")
the way WARN_ON() is implemented is suboptimal. That commit also
mentions an issue with 'long long' condition. It fixed it for
WARN_ON() but the same problem still exists today with BUG_ON() on
PPC32. It will be fixed by using the generic implementation.

By using the generic implementation, gcc will naturally generate a
branch to the unconditional trap generated by BUG().

As modern powerpc implement zero-cycle branch,
that's even more efficient.

And for the functions using WARN_ON() and its return, the test
on return from WARN_ON() is now also used for the WARN_ON() itself.

On PPC64 we don't want it because we want to be able to use CFAR
register to track how we entered the code that trapped. The CFAR
register would be clobbered by the branch.

A simple test function:

unsigned long test9w(unsigned long a, unsigned long b)
{
if (WARN_ON(!b))
return 0;
return a / b;
}

Before the patch:

0000046c <test9w>:
 46c: 7c 89 00 34  cntlzw  r9,r4
 470: 55 29 d9 7e  rlwinm  r9,r9,27,5,31
 474: 0f 09 00 00  twnei   r9,0
 478: 2c 04 00 00  cmpwi   r4,0
 47c: 41 82 00 0c  beq     488 <test9w+0x1c>
 480: 7c 63 23 96  divwu   r3,r3,r4
 484: 4e 80 00 20  blr

 488: 38 60 00 00  li      r3,0
 48c: 4e 80 00 20  blr

After the patch:

00000468 <test9w>:
 468: 2c 04 00 00  cmpwi   r4,0
 46c: 41 82 00 0c  beq     478 <test9w+0x10>
 470: 7c 63 23 96  divwu   r3,r3,r4
 474: 4e 80 00 20  blr

 478: 0f e0 00 00  twui    r0,0
 47c: 38 60 00 00  li      r3,0
 480: 4e 80 00 20  blr

So we see before the patch we need 3 instructions on the likely path
to handle the WARN_ON(). With the patch the trap goes on the unlikely
path.

See below the difference at the entry of system_call_exception where
we have several BUG_ON(), allthough less impressing.

With the patch:

00000000 <system_call_exception>:
   0: 81 6a 00 84  lwz     r11,132(r10)
   4: 90 6a 00 88  stw     r3,136(r10)
   8: 71 60 00 02  andi.   r0,r11,2
   c: 41 82 00 70  beq     7c <system_call_exception+0x7c>
  10: 71 60 40 00  andi.   r0,r11,16384
  14: 41 82 00 6c  beq     80 <system_call_exception+0x80>
  18: 71 6b 80 00  andi.   r11,r11,32768
  1c: 41 82 00 68  beq     84 <system_call_exception+0x84>
  20: 94 21 ff e0  stwu    r1,-32(r1)
  24: 93 e1 00 1c  stw     r31,28(r1)
  28: 7d 8c 42 e6  mftb    r12
...
  7c: 0f e0 00 00  twui    r0,0
  80: 0f e0 00 00  twui    r0,0
  84: 0f e0 00 00  twui    r0,0

Without the patch:

00000000 <system_call_exception>:
   0: 94 21 ff e0  stwu    r1,-32(r1)
   4: 93 e1 00 1c  stw     r31,28(r1)
   8: 90 6a 00 88  stw     r3,136(r10)
   c: 81 6a 00 84  lwz     r11,132(r10)
  10: 69 60 00 02  xori    r0,r11,2
  14: 54 00 ff fe  rlwinm  r0,r0,31,31,31
  18: 0f 00 00 00  twnei   r0,0
  1c: 69 60 40 00  xori    r0,r11,16384
  20: 54 00 97 fe  rlwinm  r0,r0,18,31,31
  24: 0f 00 00 00  twnei   r0,0
  28: 69 6b 80 00  xori    r11,r11,32768
  2c: 55 6b 8f fe  rlwinm  r11,r11,17,31,31
  30: 0f 0b 00 00  twnei   r11,0
  34: 7d 8c 42 e6  mftb    r12

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/b286e07fb771a664b631cd07a40b09c06f26e64b.1618331881.git.christophe.leroy@csgroup.eu
2 years agopowerpc/pseries: Add support for FORM2 associativity
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:22:23 +0000 (18:52 +0530)]
powerpc/pseries: Add support for FORM2 associativity

PAPR interface currently supports two different ways of communicating resource
grouping details to the OS. These are referred to as Form 0 and Form 1
associativity grouping. Form 0 is the older format and is now considered
deprecated. This patch adds another resource grouping named FORM2.

Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132223.225214-6-aneesh.kumar@linux.ibm.com
2 years agopowerpc/pseries: Add a helper for form1 cpu distance
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:22:22 +0000 (18:52 +0530)]
powerpc/pseries: Add a helper for form1 cpu distance

This helper is only used with the dispatch trace log collection.
A later patch will add Form2 affinity support and this change helps
in keeping that simpler. Also add a comment explaining we don't expect
the code to be called with FORM0

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132223.225214-5-aneesh.kumar@linux.ibm.com
2 years agopowerpc/pseries: Consolidate different NUMA distance update code paths
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:22:21 +0000 (18:52 +0530)]
powerpc/pseries: Consolidate different NUMA distance update code paths

The associativity details of the newly added resourced are collected from
the hypervisor via "ibm,configure-connector" rtas call. Update the numa
distance details of the newly added numa node after the above call.

Instead of updating NUMA distance every time we lookup a node id
from the associativity property, add helpers that can be used
during boot which does this only once. Also remove the distance
update from node id lookup helpers.

Currently, we duplicate parsing code for ibm,associativity and
ibm,associativity-lookup-arrays in the kernel. The associativity array provided
by these device tree properties are very similar and hence can use
a helper to parse the node id and numa distance details.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132223.225214-4-aneesh.kumar@linux.ibm.com
2 years agopowerpc/pseries: Rename TYPE1_AFFINITY to FORM1_AFFINITY
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:22:20 +0000 (18:52 +0530)]
powerpc/pseries: Rename TYPE1_AFFINITY to FORM1_AFFINITY

Also make related code cleanup that will allow adding FORM2_AFFINITY in
later patches. No functional change in this patch.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132223.225214-3-aneesh.kumar@linux.ibm.com
2 years agopowerpc/pseries: rename min_common_depth to primary_domain_index
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:22:19 +0000 (18:52 +0530)]
powerpc/pseries: rename min_common_depth to primary_domain_index

No functional change in this patch.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132223.225214-2-aneesh.kumar@linux.ibm.com
2 years agopowerpc: rename powerpc_debugfs_root to arch_debugfs_dir
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:28:31 +0000 (18:58 +0530)]
powerpc: rename powerpc_debugfs_root to arch_debugfs_dir

No functional change in this patch. arch_debugfs_dir is the generic kernel
name declared in linux/debugfs.h for arch-specific debugfs directory.
Architectures like x86/s390 already use the name. Rename powerpc
specific powerpc_debugfs_root to arch_debugfs_dir.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132831.233794-2-aneesh.kumar@linux.ibm.com
2 years agopowerpc/book3s64/radix: make tlb_single_page_flush_ceiling a debugfs entry
Aneesh Kumar K.V [Thu, 12 Aug 2021 13:28:30 +0000 (18:58 +0530)]
powerpc/book3s64/radix: make tlb_single_page_flush_ceiling a debugfs entry

Similar to x86/s390 add a debugfs file to tune tlb_single_page_flush_ceiling.
Also add a debugfs entry for tlb_local_single_page_flush_ceiling.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210812132831.233794-1-aneesh.kumar@linux.ibm.com
2 years agocpufreq: powernv: Fix init_chip_info initialization in numa=off
Pratik R. Sampat [Wed, 28 Jul 2021 12:05:00 +0000 (17:35 +0530)]
cpufreq: powernv: Fix init_chip_info initialization in numa=off

In the numa=off kernel command-line configuration init_chip_info() loops
around the number of chips and attempts to copy the cpumask of that node
which is NULL for all iterations after the first chip.

Hence, store the cpu mask for each chip instead of derving cpumask from
node while populating the "chips" struct array and copy that to the
chips[i].mask

Fixes: 053819e0bf84 ("cpufreq: powernv: Handle throttling due to Pmax capping at chip level")
Cc: stable@vger.kernel.org # v4.3+
Reported-by: Shirisha Ganta <shirisha.ganta1@ibm.com>
Signed-off-by: Pratik R. Sampat <psampat@linux.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
[mpe: Rename goto label to out_free_chip_cpu_mask]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210728120500.87549-2-psampat@linux.ibm.com
2 years agopowerpc: wii_defconfig: Enable OTP by default
Emmanuel Gil Peyrot [Sun, 1 Aug 2021 07:38:22 +0000 (09:38 +0200)]
powerpc: wii_defconfig: Enable OTP by default

This selects the nintendo-otp module when building for this platform.

Signed-off-by: Emmanuel Gil Peyrot <linkmauve@linkmauve.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210801073822.12452-6-linkmauve@linkmauve.fr
2 years agopowerpc: wii.dts: Expose the OTP on this platform
Emmanuel Gil Peyrot [Sun, 1 Aug 2021 07:38:21 +0000 (09:38 +0200)]
powerpc: wii.dts: Expose the OTP on this platform

This can be used by the newly-added nintendo-otp nvmem module.

Signed-off-by: Emmanuel Gil Peyrot <linkmauve@linkmauve.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210801073822.12452-5-linkmauve@linkmauve.fr
2 years agopowerpc: wii.dts: Reduce the size of the control area
Emmanuel Gil Peyrot [Sun, 1 Aug 2021 07:38:20 +0000 (09:38 +0200)]
powerpc: wii.dts: Reduce the size of the control area

This is wrong, but needed in order to avoid overlapping ranges with the
OTP area added in the next commit.  A refactor of this part of the
device tree is needed: according to Wiibrew[1], this area starts at
0x0d800000 and spans 0x400 bytes (that is, 0x100 32-bit registers),
encompassing PIC and GPIO registers, amongst the ones already exposed in
this device tree, which should become children of the control@d800000
node.

[1] https://wiibrew.org/wiki/Hardware/Hollywood_Registers

Signed-off-by: Emmanuel Gil Peyrot <linkmauve@linkmauve.fr>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210801073822.12452-4-linkmauve@linkmauve.fr
2 years agopowerpc: Bulk conversion to generic_handle_domain_irq()
Marc Zyngier [Mon, 2 Aug 2021 16:26:28 +0000 (17:26 +0100)]
powerpc: Bulk conversion to generic_handle_domain_irq()

Wherever possible, replace constructs that match either
generic_handle_irq(irq_find_mapping()) or
generic_handle_irq(irq_linear_revmap()) to a single call to
generic_handle_domain_irq().

Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210802162630.2219813-13-maz@kernel.org
2 years agoKVM: PPC: Book3S HV: XIVE: Add support for automatic save-restore
Cédric Le Goater [Tue, 20 Jul 2021 13:42:09 +0000 (15:42 +0200)]
KVM: PPC: Book3S HV: XIVE: Add support for automatic save-restore

On P10, the feature doing an automatic "save & restore" of a VCPU
interrupt context is set by default in OPAL. When a VP context is
pulled out, the state of the interrupt registers are saved by the XIVE
interrupt controller under the internal NVP structure representing the
VP. This saves a costly store/load in guest entries and exits.

If OPAL advertises the "save & restore" feature in the device tree,
it should also have set the 'H' bit in the CAM line. Check that when
vCPUs are connected to their ICP in KVM before going any further.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210720134209.256133-3-clg@kaod.org
2 years agoKVM: PPC: Book3S HV: XIVE: Add a 'flags' field
Cédric Le Goater [Tue, 20 Jul 2021 13:42:08 +0000 (15:42 +0200)]
KVM: PPC: Book3S HV: XIVE: Add a 'flags' field

Use it to hold platform specific features. P9 DD2 introduced
single-escalation support. P10 will add others.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210720134209.256133-2-clg@kaod.org
2 years agopowerpc: use IRQF_NO_DEBUG for IPIs
Cédric Le Goater [Mon, 19 Jul 2021 13:06:14 +0000 (15:06 +0200)]
powerpc: use IRQF_NO_DEBUG for IPIs

There is no need to use the lockup detector ("noirqdebug") for IPIs.
The ipistorm benchmark measures a ~10% improvement on high systems
when this flag is set.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210719130614.195886-1-clg@kaod.org
2 years agopowerpc/xive: Use XIVE domain under xmon and debugfs
Cédric Le Goater [Thu, 1 Jul 2021 13:27:49 +0000 (15:27 +0200)]
powerpc/xive: Use XIVE domain under xmon and debugfs

The default domain of the PCI/MSIs is not the XIVE domain anymore. To
list the IRQ mappings under XMON and debugfs, query the IRQ data from
the low level XIVE domain.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-32-clg@kaod.org
2 years agoKVM: PPC: Book3S HV: XICS: Fix mapping of passthrough interrupts
Cédric Le Goater [Thu, 1 Jul 2021 13:27:48 +0000 (15:27 +0200)]
KVM: PPC: Book3S HV: XICS: Fix mapping of passthrough interrupts

PCI MSIs now live in an MSI domain but the underlying calls, which
will EOI the interrupt in real mode, need an HW IRQ number mapped in
the XICS IRQ domain. Grab it there.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-31-clg@kaod.org
2 years agopowerpc/powernv/pci: Rework pnv_opal_pci_msi_eoi()
Cédric Le Goater [Thu, 1 Jul 2021 13:27:47 +0000 (15:27 +0200)]
powerpc/powernv/pci: Rework pnv_opal_pci_msi_eoi()

pnv_opal_pci_msi_eoi() is called from KVM to EOI passthrough interrupts
when in real mode. Adding MSI domain broke the hack using the
'ioda.irq_chip' field to deduce the owning PHB. Fix that by using the
IRQ chip data in the MSI domain.

The 'ioda.irq_chip' field is now unused and could be removed from the
pnv_phb struct.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-30-clg@kaod.org
2 years agopowerpc/powernv/pci: Set the IRQ chip data for P8/CXL devices
Cédric Le Goater [Thu, 1 Jul 2021 13:27:46 +0000 (15:27 +0200)]
powerpc/powernv/pci: Set the IRQ chip data for P8/CXL devices

Before MSI domains, the default IRQ chip of PHB3 MSIs was patched by
pnv_set_msi_irq_chip() with the custom EOI handler pnv_ioda2_msi_eoi()
and the owning PHB was deduced from the 'ioda.irq_chip' field. This
path has been deprecated by the MSI domains but it is still in use by
the P8 CAPI 'cxl' driver.

Rewriting this driver to support MSI would be a waste of time.
Nevertheless, we can still remove the IRQ chip patch and set the IRQ
chip data instead. This is cleaner.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-29-clg@kaod.org
2 years agopowerpc/xics: Fix IRQ migration
Cédric Le Goater [Thu, 1 Jul 2021 13:27:45 +0000 (15:27 +0200)]
powerpc/xics: Fix IRQ migration

desc->irq_data points to the top level IRQ data descriptor which is
not necessarily in the XICS IRQ domain. MSIs are in another domain for
instance. Fix that by looking for a mapping on the low level XICS IRQ
domain.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-28-clg@kaod.org
2 years agopowerpc/powernv/pci: Adapt is_pnv_opal_msi() to detect passthrough interrupt
Cédric Le Goater [Thu, 1 Jul 2021 13:27:44 +0000 (15:27 +0200)]
powerpc/powernv/pci: Adapt is_pnv_opal_msi() to detect passthrough interrupt

The pnv_ioda2_msi_eoi() chip handler is not used anymore for MSIs.
Simply use the check on the PSI-MSI chip.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-27-clg@kaod.org
2 years agopowerpc/powernv/pci: Drop unused MSI code
Cédric Le Goater [Thu, 1 Jul 2021 13:27:43 +0000 (15:27 +0200)]
powerpc/powernv/pci: Drop unused MSI code

MSIs should be fully managed by the PCI and IRQ subsystems now.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-26-clg@kaod.org
2 years agopowerpc/pseries/pci: Drop unused MSI code
Cédric Le Goater [Thu, 1 Jul 2021 13:27:42 +0000 (15:27 +0200)]
powerpc/pseries/pci: Drop unused MSI code

MSIs should be fully managed by the PCI and IRQ subsystems now.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-25-clg@kaod.org
2 years agopowerpc/xics: Drop unmask of MSIs at startup
Cédric Le Goater [Thu, 1 Jul 2021 13:27:41 +0000 (15:27 +0200)]
powerpc/xics: Drop unmask of MSIs at startup

That was a workaround in the XICS domain because of the lack of MSI
domain. This is now handled.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-24-clg@kaod.org
2 years agopowerpc/pci: Drop XIVE restriction on MSI domains
Cédric Le Goater [Thu, 1 Jul 2021 13:27:40 +0000 (15:27 +0200)]
powerpc/pci: Drop XIVE restriction on MSI domains

The PowerNV and pSeries platforms now have support for both the XICS
and XIVE IRQ domains.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-23-clg@kaod.org
2 years agopowerpc/powernv/pci: Customize the MSI EOI handler to support PHB3
Cédric Le Goater [Thu, 1 Jul 2021 13:27:39 +0000 (15:27 +0200)]
powerpc/powernv/pci: Customize the MSI EOI handler to support PHB3

PHB3s need an extra OPAL call to EOI the interrupt. The call takes an
OPAL HW IRQ number but it is translated into a vector number in OPAL.
Here, we directly use the vector number of the in-the-middle "PNV-MSI"
domain instead of grabbing the OPAL HW IRQ number in the XICS parent
domain.

Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210701132750.1475580-22-clg@kaod.org