kernel/linux.git/arch/powerpc/platforms/pseries/hotplug-cpu.c, branch linux-5.9.y

powerpc/pseries/hotplug-cpu: wait indefinitely for vCPU death

2020-08-18T07:09:52+00:00

For a power9 KVM guest with XIVE enabled, running a test loop where we hotplug 384 vcpus and then unplug them, the following traces can be seen (generally within a few loops) either from the unplugged vcpu: cpu 65 (hwid 65) Ready to die... Querying DEAD? cpu 66 (66) shows 2 list_del corruption. next->prev should be c00a000002470208, but was c00a000002470048 ------------[ cut here ]------------ kernel BUG at lib/list_debug.c:56! Oops: Exception in kernel mode, sig: 5 [#1] LE SMP NR_CPUS=2048 NUMA pSeries Modules linked in: fuse nft_fib_inet nft_fib_ipv4 nft_fib_ipv6 ... CPU: 66 PID: 0 Comm: swapper/66 Kdump: loaded Not tainted 4.18.0-221.el8.ppc64le #1 NIP: c0000000007ab50c LR: c0000000007ab508 CTR: 00000000000003ac REGS: c0000009e5a17840 TRAP: 0700 Not tainted (4.18.0-221.el8.ppc64le) MSR: 800000000282b033 CR: 28000842 XER: 20040000 ... NIP __list_del_entry_valid+0xac/0x100 LR __list_del_entry_valid+0xa8/0x100 Call Trace: __list_del_entry_valid+0xa8/0x100 (unreliable) free_pcppages_bulk+0x1f8/0x940 free_unref_page+0xd0/0x100 xive_spapr_cleanup_queue+0x148/0x1b0 xive_teardown_cpu+0x1bc/0x240 pseries_mach_cpu_die+0x78/0x2f0 cpu_die+0x48/0x70 arch_cpu_idle_dead+0x20/0x40 do_idle+0x2f4/0x4c0 cpu_startup_entry+0x38/0x40 start_secondary+0x7bc/0x8f0 start_secondary_prolog+0x10/0x14 or on the worker thread handling the unplug: pseries-hotplug-cpu: Attempting to remove CPU , drc index: 1000013a Querying DEAD? cpu 314 (314) shows 2 BUG: Bad page state in process kworker/u768:3 pfn:95de1 cpu 314 (hwid 314) Ready to die... page:c00a000002577840 refcount:0 mapcount:-128 mapping:0000000000000000 index:0x0 flags: 0x5ffffc00000000() raw: 005ffffc00000000 5deadbeef0000100 5deadbeef0000200 0000000000000000 raw: 0000000000000000 0000000000000000 00000000ffffff7f 0000000000000000 page dumped because: nonzero mapcount Modules linked in: kvm xt_CHECKSUM ipt_MASQUERADE xt_conntrack ... CPU: 0 PID: 548 Comm: kworker/u768:3 Kdump: loaded Not tainted 4.18.0-224.el8.bz1856588.ppc64le #1 Workqueue: pseries hotplug workque pseries_hp_work_fn Call Trace: dump_stack+0xb0/0xf4 (unreliable) bad_page+0x12c/0x1b0 free_pcppages_bulk+0x5bc/0x940 page_alloc_cpu_dead+0x118/0x120 cpuhp_invoke_callback.constprop.5+0xb8/0x760 _cpu_down+0x188/0x340 cpu_down+0x5c/0xa0 cpu_subsys_offline+0x24/0x40 device_offline+0xf0/0x130 dlpar_offline_cpu+0x1c4/0x2a0 dlpar_cpu_remove+0xb8/0x190 dlpar_cpu_remove_by_index+0x12c/0x150 dlpar_cpu+0x94/0x800 pseries_hp_work_fn+0x128/0x1e0 process_one_work+0x304/0x5d0 worker_thread+0xcc/0x7a0 kthread+0x1ac/0x1c0 ret_from_kernel_thread+0x5c/0x80 The latter trace is due to the following sequence: page_alloc_cpu_dead drain_pages drain_pages_zone free_pcppages_bulk where drain_pages() in this case is called under the assumption that the unplugged cpu is no longer executing. To ensure that is the case, and early call is made to __cpu_die()->pseries_cpu_die(), which runs a loop that waits for the cpu to reach a halted state by polling its status via query-cpu-stopped-state RTAS calls. It only polls for 25 iterations before giving up, however, and in the trace above this results in the following being printed only .1 seconds after the hotplug worker thread begins processing the unplug request: pseries-hotplug-cpu: Attempting to remove CPU , drc index: 1000013a Querying DEAD? cpu 314 (314) shows 2 At that point the worker thread assumes the unplugged CPU is in some unknown/dead state and procedes with the cleanup, causing the race with the XIVE cleanup code executed by the unplugged CPU. Fix this by waiting indefinitely, but also making an effort to avoid spurious lockup messages by allowing for rescheduling after polling the CPU status and printing a warning if we wait for longer than 120s. Fixes: eac1e731b59ee ("powerpc/xive: guest exploitation of the XIVE interrupt controller") Suggested-by: Michael Ellerman Signed-off-by: Michael Roth Tested-by: Greg Kurz Reviewed-by: Thiago Jung Bauermann Reviewed-by: Greg Kurz [mpe: Trim oopses in change log slightly for readability] Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20200811161544.10513-1-mdroth@linux.vnet.ibm.com

powerpc/pseries/hotplug-cpu: Remove double free in error path

2020-07-30T12:53:49+00:00

In the unlikely event that the device tree lacks a /cpus node, find_dlpar_cpus_to_add() oddly frees the cpu_drcs buffer it has been passed before returning an error. Its only caller also frees the buffer on error. Remove the less conventional kfree() of a caller-supplied buffer from find_dlpar_cpus_to_add(). Fixes: 90edf184b9b7 ("powerpc/pseries: Add CPU dlpar add functionality") Signed-off-by: Nathan Lynch Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20190919231633.1344-1-nathanl@linux.ibm.com

powerpc/pseries: remove dlpar_cpu_readd()

2020-07-16T03:12:40+00:00

dlpar_cpu_readd() is unused now. Signed-off-by: Nathan Lynch Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20200612051238.1007764-18-nathanl@linux.ibm.com

powerpc/numa: remove timed_topology_update()

2020-07-16T03:12:37+00:00

timed_topology_update is a no-op now, so remove it and all call sites. Signed-off-by: Nathan Lynch Reviewed-by: Srikar Dronamraju Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20200612051238.1007764-11-nathanl@linux.ibm.com

powerpc/pseries: remove cede offline state for CPUs

2020-07-16T03:12:34+00:00

This effectively reverts commit 3aa565f53c39 ("powerpc/pseries: Add hooks to put the CPU into an appropriate offline state"), which added an offline mode for CPUs which uses the H_CEDE hcall instead of the architected stop-self RTAS function in order to facilitate "folding" of dedicated mode processors on PowerVM platforms to achieve energy savings. This has been the default offline mode since its introduction. There's nothing about stop-self that would prevent the hypervisor from achieving the energy savings available via H_CEDE, so the original premise of this change appears to be flawed. I also have encountered the claim that the transition to and from ceded state is much faster than stop-self/start-cpu. Certainly we would not want to use stop-self as an *idle* mode. That is what H_CEDE is for. However, this difference is insignificant in the context of Linux CPU hotplug, where the latency of an offline or online operation on current systems is on the order of 100ms, mainly attributable to all the various subsystems' cpuhp callbacks. The cede offline mode also prevents accurate accounting, as discussed before: https://lore.kernel.org/linuxppc-dev/1571740391-3251-1-git-send-email-ego@linux.vnet.ibm.com/ Unconditionally use stop-self to offline processor threads. This is the architected method for offlining CPUs on PAPR systems. The "cede_offline" boot parameter is rendered obsolete. Removing this code enables the removal of the partition suspend code which temporarily onlines all present CPUs. Fixes: 3aa565f53c39 ("powerpc/pseries: Add hooks to put the CPU into an appropriate offline state") Signed-off-by: Nathan Lynch Reviewed-by: Gautham R. Shenoy Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20200612051238.1007764-2-nathanl@linux.ibm.com

powerpc/pseries: Add cpu DLPAR support for drc-info property

2019-11-13T05:57:57+00:00

Older firmwares provided information about Dynamic Reconfig Connectors (DRC) through several device tree properties, namely ibm,drc-types, ibm,drc-indexes, ibm,drc-names, and ibm,drc-power-domains. New firmwares have the ability to present this same information in a much condensed format through a device tree property called ibm,drc-info. The existing cpu DLPAR hotplug code only understands the older DRC property format when validating the drc-index of a cpu during a hotplug add. This updates those code paths to use the ibm,drc-info property, when present, instead for validation. Signed-off-by: Tyrel Datwyler Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/1573449697-5448-4-git-send-email-tyreld@linux.ibm.com

powerpc/pseries: safely roll back failed DLPAR cpu add

2019-11-05T00:26:45+00:00

dlpar_online_cpu() attempts to online all threads of a core that has been added to an LPAR. If onlining a non-primary thread fails (e.g. due to an allocation failure), the core is left with at least one thread online. dlpar_cpu_add() attempts to roll back the whole operation, releasing the core back to the platform. However, since some threads of the core being removed are still online, the BUG_ON(cpu_online(cpu)) in pseries_remove_processor() strikes: LE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=2048 NUMA pSeries Modules linked in: CPU: 3 PID: 8587 Comm: drmgr Not tainted 5.3.0-rc2-00190-g9b123d1ea237-dirty #46 NIP: c0000000000eeb2c LR: c0000000000eeac4 CTR: c0000000000ee9e0 REGS: c0000001f745b6c0 TRAP: 0700 Not tainted (5.3.0-rc2-00190-g9b123d1ea237-dirty) MSR: 800000010282b033 CR: 44002448 XER: 00000000 CFAR: c00000000195d718 IRQMASK: 0 GPR00: c0000000000eeac4 c0000001f745b950 c0000000032f6200 0000000000000008 GPR04: 0000000000000008 c000000003349c78 0000000000000040 00000000000001ff GPR08: 0000000000000008 0000000000000000 0000000000000001 0007ffffffffffff GPR12: 0000000084002844 c00000001ecacb80 0000000000000000 0000000000000000 GPR16: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 GPR20: 0000000000000000 0000000000000000 0000000000000000 0000000000000008 GPR24: c000000003349ee0 c00000000334a2e4 c0000000fca4d7a8 c000000001d20048 GPR28: 0000000000000001 ffffffffffffffff ffffffffffffffff c0000000fca4d7c4 NIP [c0000000000eeb2c] pseries_smp_notifier+0x14c/0x2e0 LR [c0000000000eeac4] pseries_smp_notifier+0xe4/0x2e0 Call Trace: [c0000001f745b950] [c0000000000eeac4] pseries_smp_notifier+0xe4/0x2e0 (unreliable) [c0000001f745ba10] [c0000000001ac774] notifier_call_chain+0xb4/0x190 [c0000001f745bab0] [c0000000001ad62c] blocking_notifier_call_chain+0x7c/0xb0 [c0000001f745baf0] [c00000000167bda0] of_detach_node+0xc0/0x110 [c0000001f745bb50] [c0000000000e7ae4] dlpar_detach_node+0x64/0xa0 [c0000001f745bb80] [c0000000000edefc] dlpar_cpu_add+0x31c/0x360 [c0000001f745bc10] [c0000000000ee980] dlpar_cpu_probe+0x50/0xb0 [c0000001f745bc50] [c00000000002cf70] arch_cpu_probe+0x40/0x70 [c0000001f745bc70] [c000000000ccd808] cpu_probe_store+0x48/0x80 [c0000001f745bcb0] [c000000000cbcef8] dev_attr_store+0x38/0x60 [c0000001f745bcd0] [c00000000059c980] sysfs_kf_write+0x70/0xb0 [c0000001f745bd10] [c00000000059afb8] kernfs_fop_write+0xf8/0x280 [c0000001f745bd60] [c0000000004b437c] __vfs_write+0x3c/0x70 [c0000001f745bd80] [c0000000004b8710] vfs_write+0xd0/0x220 [c0000001f745bdd0] [c0000000004b8acc] ksys_write+0x7c/0x140 [c0000001f745be20] [c00000000000bbd8] system_call+0x5c/0x68 Move dlpar_offline_cpu() up in the file so that dlpar_online_cpu() can use it to re-offline any threads that have been onlined when an error is encountered. Signed-off-by: Nathan Lynch Fixes: e666ae0b10aa ("powerpc/pseries: Update CPU hotplug error recovery") Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20191016183611.10867-3-nathanl@linux.ibm.com

powerpc/pseries: address checkpatch warnings in dlpar_offline_cpu

2019-11-05T00:26:45+00:00

Remove some stray blank lines, convert a printk to pr_warn, and address a line length violation. One functional change: use WARN_ON instead of BUG_ON in case H_PROD of a ceded thread yields an unexpected result from the platform. We can expect this code path to get uninterruptibly stuck in __cpu_die() if this happens, but that's more desirable than crashing. Signed-off-by: Nathan Lynch Fixes: b6db63d1a7f0 ("pseries/pseries: Add code to online/offline CPUs of a DLPAR node") Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20191016183611.10867-2-nathanl@linux.ibm.com

treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 152

2019-05-30T18:26:32+00:00

Based on 1 normalized pattern(s): this program is free software you can redistribute it and or modify it under the terms of the gnu general public license as published by the free software foundation either version 2 of the license or at your option any later version extracted by the scancode license scanner the SPDX license identifier GPL-2.0-or-later has been chosen to replace the boilerplate/reference in 3029 file(s). Signed-off-by: Thomas Gleixner Reviewed-by: Allison Randal Cc: linux-spdx@vger.kernel.org Link: https://lkml.kernel.org/r/20190527070032.746973796@linutronix.de Signed-off-by: Greg Kroah-Hartman

powerpc/pseries: Perform full re-add of CPU for topology update post-migration

2019-01-30T12:28:56+00:00

On pseries systems, performing a partition migration can result in altering the nodes a CPU is assigned to on the destination system. For exampl, pre-migration on the source system CPUs are in node 1 and 3, post-migration on the destination system CPUs are in nodes 2 and 3. Handling the node change for a CPU can cause corruption in the slab cache if we hit a timing where a CPUs node is changed while cache_reap() is invoked. The corruption occurs because the slab cache code appears to rely on the CPU and slab cache pages being on the same node. The current dynamic updating of a CPUs node done in arch/powerpc/mm/numa.c does not prevent us from hitting this scenario. Changing the device tree property update notification handler that recognizes an affinity change for a CPU to do a full DLPAR remove and add of the CPU instead of dynamically changing its node resolves this issue. Signed-off-by: Nathan Fontenot Signed-off-by: Michael W. Bringmann Tested-by: Michael W. Bringmann Signed-off-by: Michael Ellerman