From 3a83f677a6eeff65751b29e3648d7c69c3be83f3 Mon Sep 17 00:00:00 2001 From: Michael Roth Date: Wed, 11 Sep 2019 17:31:55 -0500 Subject: KVM: PPC: Book3S HV: use smp_mb() when setting/clearing host_ipi flag On a 2-socket Power9 system with 32 cores/128 threads (SMT4) and 1TB of memory running the following guest configs: guest A: - 224GB of memory - 56 VCPUs (sockets=1,cores=28,threads=2), where: VCPUs 0-1 are pinned to CPUs 0-3, VCPUs 2-3 are pinned to CPUs 4-7, ... VCPUs 54-55 are pinned to CPUs 108-111 guest B: - 4GB of memory - 4 VCPUs (sockets=1,cores=4,threads=1) with the following workloads (with KSM and THP enabled in all): guest A: stress --cpu 40 --io 20 --vm 20 --vm-bytes 512M guest B: stress --cpu 4 --io 4 --vm 4 --vm-bytes 512M host: stress --cpu 4 --io 4 --vm 2 --vm-bytes 256M the below soft-lockup traces were observed after an hour or so and persisted until the host was reset (this was found to be reliably reproducible for this configuration, for kernels 4.15, 4.18, 5.0, and 5.3-rc5): [ 1253.183290] rcu: INFO: rcu_sched self-detected stall on CPU [ 1253.183319] rcu: 124-....: (5250 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=1941 [ 1256.287426] watchdog: BUG: soft lockup - CPU#105 stuck for 23s! [CPU 52/KVM:19709] [ 1264.075773] watchdog: BUG: soft lockup - CPU#24 stuck for 23s! [worker:19913] [ 1264.079769] watchdog: BUG: soft lockup - CPU#31 stuck for 23s! [worker:20331] [ 1264.095770] watchdog: BUG: soft lockup - CPU#45 stuck for 23s! [worker:20338] [ 1264.131773] watchdog: BUG: soft lockup - CPU#64 stuck for 23s! [avocado:19525] [ 1280.408480] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791] [ 1316.198012] rcu: INFO: rcu_sched self-detected stall on CPU [ 1316.198032] rcu: 124-....: (21003 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=8243 [ 1340.411024] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791] [ 1379.212609] rcu: INFO: rcu_sched self-detected stall on CPU [ 1379.212629] rcu: 124-....: (36756 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=14714 [ 1404.413615] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791] [ 1442.227095] rcu: INFO: rcu_sched self-detected stall on CPU [ 1442.227115] rcu: 124-....: (52509 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=21403 [ 1455.111787] INFO: task worker:19907 blocked for more than 120 seconds. [ 1455.111822] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.111833] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.111884] INFO: task worker:19908 blocked for more than 120 seconds. [ 1455.111905] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.111925] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.111966] INFO: task worker:20328 blocked for more than 120 seconds. [ 1455.111986] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.111998] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.112048] INFO: task worker:20330 blocked for more than 120 seconds. [ 1455.112068] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.112097] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.112138] INFO: task worker:20332 blocked for more than 120 seconds. [ 1455.112159] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.112179] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.112210] INFO: task worker:20333 blocked for more than 120 seconds. [ 1455.112231] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.112242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.112282] INFO: task worker:20335 blocked for more than 120 seconds. [ 1455.112303] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 [ 1455.112332] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 1455.112372] INFO: task worker:20336 blocked for more than 120 seconds. [ 1455.112392] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1 CPUs 45, 24, and 124 are stuck on spin locks, likely held by CPUs 105 and 31. CPUs 105 and 31 are stuck in smp_call_function_many(), waiting on target CPU 42. For instance: # CPU 105 registers (via xmon) R00 = c00000000020b20c R16 = 00007d1bcd800000 R01 = c00000363eaa7970 R17 = 0000000000000001 R02 = c0000000019b3a00 R18 = 000000000000006b R03 = 000000000000002a R19 = 00007d537d7aecf0 R04 = 000000000000002a R20 = 60000000000000e0 R05 = 000000000000002a R21 = 0801000000000080 R06 = c0002073fb0caa08 R22 = 0000000000000d60 R07 = c0000000019ddd78 R23 = 0000000000000001 R08 = 000000000000002a R24 = c00000000147a700 R09 = 0000000000000001 R25 = c0002073fb0ca908 R10 = c000008ffeb4e660 R26 = 0000000000000000 R11 = c0002073fb0ca900 R27 = c0000000019e2464 R12 = c000000000050790 R28 = c0000000000812b0 R13 = c000207fff623e00 R29 = c0002073fb0ca808 R14 = 00007d1bbee00000 R30 = c0002073fb0ca800 R15 = 00007d1bcd600000 R31 = 0000000000000800 pc = c00000000020b260 smp_call_function_many+0x3d0/0x460 cfar= c00000000020b270 smp_call_function_many+0x3e0/0x460 lr = c00000000020b20c smp_call_function_many+0x37c/0x460 msr = 900000010288b033 cr = 44024824 ctr = c000000000050790 xer = 0000000000000000 trap = 100 CPU 42 is running normally, doing VCPU work: # CPU 42 stack trace (via xmon) [link register ] c00800001be17188 kvmppc_book3s_radix_page_fault+0x90/0x2b0 [kvm_hv] [c000008ed3343820] c000008ed3343850 (unreliable) [c000008ed33438d0] c00800001be11b6c kvmppc_book3s_hv_page_fault+0x264/0xe30 [kvm_hv] [c000008ed33439d0] c00800001be0d7b4 kvmppc_vcpu_run_hv+0x8dc/0xb50 [kvm_hv] [c000008ed3343ae0] c00800001c10891c kvmppc_vcpu_run+0x34/0x48 [kvm] [c000008ed3343b00] c00800001c10475c kvm_arch_vcpu_ioctl_run+0x244/0x420 [kvm] [c000008ed3343b90] c00800001c0f5a78 kvm_vcpu_ioctl+0x470/0x7c8 [kvm] [c000008ed3343d00] c000000000475450 do_vfs_ioctl+0xe0/0xc70 [c000008ed3343db0] c0000000004760e4 ksys_ioctl+0x104/0x120 [c000008ed3343e00] c000000000476128 sys_ioctl+0x28/0x80 [c000008ed3343e20] c00000000000b388 system_call+0x5c/0x70 --- Exception: c00 (System Call) at 00007d545cfd7694 SP (7d53ff7edf50) is in userspace It was subsequently found that ipi_message[PPC_MSG_CALL_FUNCTION] was set for CPU 42 by at least 1 of the CPUs waiting in smp_call_function_many(), but somehow the corresponding call_single_queue entries were never processed by CPU 42, causing the callers to spin in csd_lock_wait() indefinitely. Nick Piggin suggested something similar to the following sequence as a possible explanation (interleaving of CALL_FUNCTION/RESCHEDULE IPI messages seems to be most common, but any mix of CALL_FUNCTION and !CALL_FUNCTION messages could trigger it): CPU X: smp_muxed_ipi_set_message(): X: smp_mb() X: message[RESCHEDULE] = 1 X: doorbell_global_ipi(42): X: kvmppc_set_host_ipi(42, 1) X: ppc_msgsnd_sync()/smp_mb() X: ppc_msgsnd() -> 42 42: doorbell_exception(): // from CPU X 42: ppc_msgsync() 105: smp_muxed_ipi_set_message(): 105: smb_mb() // STORE DEFERRED DUE TO RE-ORDERING --105: message[CALL_FUNCTION] = 1 | 105: doorbell_global_ipi(42): | 105: kvmppc_set_host_ipi(42, 1) | 42: kvmppc_set_host_ipi(42, 0) | 42: smp_ipi_demux_relaxed() | 42: // returns to executing guest | // RE-ORDERED STORE COMPLETES ->105: message[CALL_FUNCTION] = 1 105: ppc_msgsnd_sync()/smp_mb() 105: ppc_msgsnd() -> 42 42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored 105: // hangs waiting on 42 to process messages/call_single_queue This can be prevented with an smp_mb() at the beginning of kvmppc_set_host_ipi(), such that stores to message[] (or other state indicated by the host_ipi flag) are ordered vs. the store to to host_ipi. However, doing so might still allow for the following scenario (not yet observed): CPU X: smp_muxed_ipi_set_message(): X: smp_mb() X: message[RESCHEDULE] = 1 X: doorbell_global_ipi(42): X: kvmppc_set_host_ipi(42, 1) X: ppc_msgsnd_sync()/smp_mb() X: ppc_msgsnd() -> 42 42: doorbell_exception(): // from CPU X 42: ppc_msgsync() // STORE DEFERRED DUE TO RE-ORDERING -- 42: kvmppc_set_host_ipi(42, 0) | 42: smp_ipi_demux_relaxed() | 105: smp_muxed_ipi_set_message(): | 105: smb_mb() | 105: message[CALL_FUNCTION] = 1 | 105: doorbell_global_ipi(42): | 105: kvmppc_set_host_ipi(42, 1) | // RE-ORDERED STORE COMPLETES -> 42: kvmppc_set_host_ipi(42, 0) 42: // returns to executing guest 105: ppc_msgsnd_sync()/smp_mb() 105: ppc_msgsnd() -> 42 42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored 105: // hangs waiting on 42 to process messages/call_single_queue Fixing this scenario would require an smp_mb() *after* clearing host_ipi flag in kvmppc_set_host_ipi() to order the store vs. subsequent processing of IPI messages. To handle both cases, this patch splits kvmppc_set_host_ipi() into separate set/clear functions, where we execute smp_mb() prior to setting host_ipi flag, and after clearing host_ipi flag. These functions pair with each other to synchronize the sender and receiver sides. With that change in place the above workload ran for 20 hours without triggering any lock-ups. Fixes: 755563bc79c7 ("powerpc/powernv: Fixes for hypervisor doorbell handling") # v4.0 Signed-off-by: Michael Roth Acked-by: Paul Mackerras Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20190911223155.16045-1-mdroth@linux.vnet.ibm.com --- arch/powerpc/platforms/powernv/smp.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch/powerpc/platforms') diff --git a/arch/powerpc/platforms/powernv/smp.c b/arch/powerpc/platforms/powernv/smp.c index 94cd96b9b7bb..fbd6e6b7bbf2 100644 --- a/arch/powerpc/platforms/powernv/smp.c +++ b/arch/powerpc/platforms/powernv/smp.c @@ -193,7 +193,7 @@ static void pnv_smp_cpu_kill_self(void) * for coming online, which are handled via * generic_check_cpu_restart() calls. */ - kvmppc_set_host_ipi(cpu, 0); + kvmppc_clear_host_ipi(cpu); srr1 = pnv_cpu_offline(cpu); -- cgit v1.2.3 From 1211ee61b4a8e60d6dc77211cdcf01906915bfba Mon Sep 17 00:00:00 2001 From: Laurent Dufour Date: Fri, 20 Sep 2019 15:05:22 +0200 Subject: powerpc/pseries: Read TLB Block Invalidate Characteristics The PAPR document specifies the TLB Block Invalidate Characteristics which tells for each pair of segment base page size, actual page size, the size of the block the hcall H_BLOCK_REMOVE supports. These characteristics are loaded at boot time in a new table hblkr_size. The table is separate from the mmu_psize_def because this is specific to the pseries platform. A new init function, pseries_lpar_read_hblkrm_characteristics() is added to read the characteristics. It is called from pSeries_setup_arch(). Fixes: ba2dd8a26baa ("powerpc/pseries/mm: call H_BLOCK_REMOVE") Signed-off-by: Laurent Dufour Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20190920130523.20441-2-ldufour@linux.ibm.com --- arch/powerpc/platforms/pseries/lpar.c | 140 +++++++++++++++++++++++++++++++ arch/powerpc/platforms/pseries/pseries.h | 1 + arch/powerpc/platforms/pseries/setup.c | 1 + 3 files changed, 142 insertions(+) (limited to 'arch/powerpc/platforms') diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c index 36b846f6e74e..8e3ac7ae4e11 100644 --- a/arch/powerpc/platforms/pseries/lpar.c +++ b/arch/powerpc/platforms/pseries/lpar.c @@ -56,6 +56,15 @@ EXPORT_SYMBOL(plpar_hcall); EXPORT_SYMBOL(plpar_hcall9); EXPORT_SYMBOL(plpar_hcall_norets); +/* + * H_BLOCK_REMOVE supported block size for this page size in segment who's base + * page size is that page size. + * + * The first index is the segment base page size, the second one is the actual + * page size. + */ +static int hblkrm_size[MMU_PAGE_COUNT][MMU_PAGE_COUNT] __ro_after_init; + #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE static u8 dtl_mask = DTL_LOG_PREEMPT; #else @@ -1311,6 +1320,137 @@ static void do_block_remove(unsigned long number, struct ppc64_tlb_batch *batch, (void)call_block_remove(pix, param, true); } +/* + * TLB Block Invalidate Characteristics + * + * These characteristics define the size of the block the hcall H_BLOCK_REMOVE + * is able to process for each couple segment base page size, actual page size. + * + * The ibm,get-system-parameter properties is returning a buffer with the + * following layout: + * + * [ 2 bytes size of the RTAS buffer (excluding these 2 bytes) ] + * ----------------- + * TLB Block Invalidate Specifiers: + * [ 1 byte LOG base 2 of the TLB invalidate block size being specified ] + * [ 1 byte Number of page sizes (N) that are supported for the specified + * TLB invalidate block size ] + * [ 1 byte Encoded segment base page size and actual page size + * MSB=0 means 4k segment base page size and actual page size + * MSB=1 the penc value in mmu_psize_def ] + * ... + * ----------------- + * Next TLB Block Invalidate Specifiers... + * ----------------- + * [ 0 ] + */ +static inline void set_hblkrm_bloc_size(int bpsize, int psize, + unsigned int block_size) +{ + if (block_size > hblkrm_size[bpsize][psize]) + hblkrm_size[bpsize][psize] = block_size; +} + +/* + * Decode the Encoded segment base page size and actual page size. + * PAPR specifies: + * - bit 7 is the L bit + * - bits 0-5 are the penc value + * If the L bit is 0, this means 4K segment base page size and actual page size + * otherwise the penc value should be read. + */ +#define HBLKRM_L_MASK 0x80 +#define HBLKRM_PENC_MASK 0x3f +static inline void __init check_lp_set_hblkrm(unsigned int lp, + unsigned int block_size) +{ + unsigned int bpsize, psize; + + /* First, check the L bit, if not set, this means 4K */ + if ((lp & HBLKRM_L_MASK) == 0) { + set_hblkrm_bloc_size(MMU_PAGE_4K, MMU_PAGE_4K, block_size); + return; + } + + lp &= HBLKRM_PENC_MASK; + for (bpsize = 0; bpsize < MMU_PAGE_COUNT; bpsize++) { + struct mmu_psize_def *def = &mmu_psize_defs[bpsize]; + + for (psize = 0; psize < MMU_PAGE_COUNT; psize++) { + if (def->penc[psize] == lp) { + set_hblkrm_bloc_size(bpsize, psize, block_size); + return; + } + } + } +} + +#define SPLPAR_TLB_BIC_TOKEN 50 + +/* + * The size of the TLB Block Invalidate Characteristics is variable. But at the + * maximum it will be the number of possible page sizes *2 + 10 bytes. + * Currently MMU_PAGE_COUNT is 16, which means 42 bytes. Use a cache line size + * (128 bytes) for the buffer to get plenty of space. + */ +#define SPLPAR_TLB_BIC_MAXLENGTH 128 + +void __init pseries_lpar_read_hblkrm_characteristics(void) +{ + unsigned char local_buffer[SPLPAR_TLB_BIC_MAXLENGTH]; + int call_status, len, idx, bpsize; + + spin_lock(&rtas_data_buf_lock); + memset(rtas_data_buf, 0, RTAS_DATA_BUF_SIZE); + call_status = rtas_call(rtas_token("ibm,get-system-parameter"), 3, 1, + NULL, + SPLPAR_TLB_BIC_TOKEN, + __pa(rtas_data_buf), + RTAS_DATA_BUF_SIZE); + memcpy(local_buffer, rtas_data_buf, SPLPAR_TLB_BIC_MAXLENGTH); + local_buffer[SPLPAR_TLB_BIC_MAXLENGTH - 1] = '\0'; + spin_unlock(&rtas_data_buf_lock); + + if (call_status != 0) { + pr_warn("%s %s Error calling get-system-parameter (0x%x)\n", + __FILE__, __func__, call_status); + return; + } + + /* + * The first two (2) bytes of the data in the buffer are the length of + * the returned data, not counting these first two (2) bytes. + */ + len = be16_to_cpu(*((u16 *)local_buffer)) + 2; + if (len > SPLPAR_TLB_BIC_MAXLENGTH) { + pr_warn("%s too large returned buffer %d", __func__, len); + return; + } + + idx = 2; + while (idx < len) { + u8 block_shift = local_buffer[idx++]; + u32 block_size; + unsigned int npsize; + + if (!block_shift) + break; + + block_size = 1 << block_shift; + + for (npsize = local_buffer[idx++]; + npsize > 0 && idx < len; npsize--) + check_lp_set_hblkrm((unsigned int) local_buffer[idx++], + block_size); + } + + for (bpsize = 0; bpsize < MMU_PAGE_COUNT; bpsize++) + for (idx = 0; idx < MMU_PAGE_COUNT; idx++) + if (hblkrm_size[bpsize][idx]) + pr_info("H_BLOCK_REMOVE supports base psize:%d psize:%d block size:%d", + bpsize, idx, hblkrm_size[bpsize][idx]); +} + /* * Take a spinlock around flushes to avoid bouncing the hypervisor tlbie * lock. diff --git a/arch/powerpc/platforms/pseries/pseries.h b/arch/powerpc/platforms/pseries/pseries.h index a6624d4bd9d0..13fa370a87e4 100644 --- a/arch/powerpc/platforms/pseries/pseries.h +++ b/arch/powerpc/platforms/pseries/pseries.h @@ -112,5 +112,6 @@ static inline unsigned long cmo_get_page_size(void) int dlpar_workqueue_init(void); void pseries_setup_rfi_flush(void); +void pseries_lpar_read_hblkrm_characteristics(void); #endif /* _PSERIES_PSERIES_H */ diff --git a/arch/powerpc/platforms/pseries/setup.c b/arch/powerpc/platforms/pseries/setup.c index f8adcd0e4589..0a40201f315f 100644 --- a/arch/powerpc/platforms/pseries/setup.c +++ b/arch/powerpc/platforms/pseries/setup.c @@ -744,6 +744,7 @@ static void __init pSeries_setup_arch(void) pseries_setup_rfi_flush(); setup_stf_barrier(); + pseries_lpar_read_hblkrm_characteristics(); /* By default, only probe PCI (can be overridden by rtas_pci) */ pci_add_flags(PCI_PROBE_ONLY); -- cgit v1.2.3 From 59545ebe331917afcb1ebbd6f3e5dc3ed51beb05 Mon Sep 17 00:00:00 2001 From: Laurent Dufour Date: Fri, 20 Sep 2019 15:05:23 +0200 Subject: powerpc/pseries: Call H_BLOCK_REMOVE when supported Depending on the hardware and the hypervisor, the hcall H_BLOCK_REMOVE may not be able to process all the page sizes for a segment base page size, as reported by the TLB Invalidate Characteristics. For each pair of base segment page size and actual page size, this characteristic tells us the size of the block the hcall supports. In the case, the hcall is not supporting a pair of base segment page size, actual page size, it is returning H_PARAM which leads to a panic like this: kernel BUG at /home/srikar/work/linux.git/arch/powerpc/platforms/pseries/lpar.c:466! Oops: Exception in kernel mode, sig: 5 [#1] BE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=2048 NUMA pSeries Modules linked in: CPU: 28 PID: 583 Comm: modprobe Not tainted 5.2.0-master #5 NIP: c0000000000be8dc LR: c0000000000be880 CTR: 0000000000000000 REGS: c0000007e77fb130 TRAP: 0700 Not tainted (5.2.0-master) MSR: 8000000000029032 CR: 42224824 XER: 20000000 CFAR: c0000000000be8fc IRQMASK: 0 GPR00: 0000000022224828 c0000007e77fb3c0 c000000001434d00 0000000000000005 GPR04: 9000000004fa8c00 0000000000000000 0000000000000003 0000000000000001 GPR08: c0000007e77fb450 0000000000000000 0000000000000001 ffffffffffffffff GPR12: c0000007e77fb450 c00000000edfcb80 0000cd7d3ea30000 c0000000016022b0 GPR16: 00000000000000b0 0000cd7d3ea30000 0000000000000001 c080001f04f00105 GPR20: 0000000000000003 0000000000000004 c000000fbeb05f58 c000000001602200 GPR24: 0000000000000000 0000000000000004 8800000000000000 c000000000c5d148 GPR28: c000000000000000 8000000000000000 a000000000000000 c0000007e77fb580 NIP [c0000000000be8dc] .call_block_remove+0x12c/0x220 LR [c0000000000be880] .call_block_remove+0xd0/0x220 Call Trace: 0xc000000fb8c00240 (unreliable) .pSeries_lpar_flush_hash_range+0x578/0x670 .flush_hash_range+0x44/0x100 .__flush_tlb_pending+0x3c/0xc0 .zap_pte_range+0x7ec/0x830 .unmap_page_range+0x3f4/0x540 .unmap_vmas+0x94/0x120 .exit_mmap+0xac/0x1f0 .mmput+0x9c/0x1f0 .do_exit+0x388/0xd60 .do_group_exit+0x54/0x100 .__se_sys_exit_group+0x14/0x20 system_call+0x5c/0x70 Instruction dump: 39400001 38a00000 4800003c 60000000 60420000 7fa9e800 38e00000 419e0014 7d29d278 7d290074 7929d182 69270001 <0b070000> 7d495378 394a0001 7fa93040 The call to H_BLOCK_REMOVE should only be made for the supported pair of base segment page size, actual page size and using the correct maximum block size. Due to the required complexity in do_block_remove() and call_block_remove(), and the fact that currently a block size of 8 is returned by the hypervisor, we are only supporting 8 size block to the H_BLOCK_REMOVE hcall. In order to identify this limitation easily in the code, a local define HBLKR_SUPPORTED_SIZE defining the currently supported block size, and a dedicated checking helper is_supported_hlbkr() are introduced. For regular pages and hugetlb, the assumption is made that the page size is equal to the base page size. For THP the page size is assumed to be 16M. Fixes: ba2dd8a26baa ("powerpc/pseries/mm: call H_BLOCK_REMOVE") Signed-off-by: Laurent Dufour Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20190920130523.20441-3-ldufour@linux.ibm.com --- arch/powerpc/platforms/pseries/lpar.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) (limited to 'arch/powerpc/platforms') diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c index 8e3ac7ae4e11..b53359258d99 100644 --- a/arch/powerpc/platforms/pseries/lpar.c +++ b/arch/powerpc/platforms/pseries/lpar.c @@ -65,6 +65,13 @@ EXPORT_SYMBOL(plpar_hcall_norets); */ static int hblkrm_size[MMU_PAGE_COUNT][MMU_PAGE_COUNT] __ro_after_init; +/* + * Due to the involved complexity, and that the current hypervisor is only + * returning this value or 0, we are limiting the support of the H_BLOCK_REMOVE + * buffer size to 8 size block. + */ +#define HBLKRM_SUPPORTED_BLOCK_SIZE 8 + #ifdef CONFIG_VIRT_CPU_ACCOUNTING_NATIVE static u8 dtl_mask = DTL_LOG_PREEMPT; #else @@ -993,6 +1000,17 @@ static void pSeries_lpar_hpte_invalidate(unsigned long slot, unsigned long vpn, #define HBLKR_CTRL_ERRNOTFOUND 0x8800000000000000UL #define HBLKR_CTRL_ERRBUSY 0xa000000000000000UL +/* + * Returned true if we are supporting this block size for the specified segment + * base page size and actual page size. + * + * Currently, we only support 8 size block. + */ +static inline bool is_supported_hlbkrm(int bpsize, int psize) +{ + return (hblkrm_size[bpsize][psize] == HBLKRM_SUPPORTED_BLOCK_SIZE); +} + /** * H_BLOCK_REMOVE caller. * @idx should point to the latest @param entry set with a PTEX. @@ -1152,7 +1170,8 @@ static inline void __pSeries_lpar_hugepage_invalidate(unsigned long *slot, if (lock_tlbie) spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags); - if (firmware_has_feature(FW_FEATURE_BLOCK_REMOVE)) + /* Assuming THP size is 16M */ + if (is_supported_hlbkrm(psize, MMU_PAGE_16M)) hugepage_block_invalidate(slot, vpn, count, psize, ssize); else hugepage_bulk_invalidate(slot, vpn, count, psize, ssize); @@ -1470,7 +1489,7 @@ static void pSeries_lpar_flush_hash_range(unsigned long number, int local) if (lock_tlbie) spin_lock_irqsave(&pSeries_lpar_tlbie_lock, flags); - if (firmware_has_feature(FW_FEATURE_BLOCK_REMOVE)) { + if (is_supported_hlbkrm(batch->psize, batch->psize)) { do_block_remove(number, batch, param); goto out; } -- cgit v1.2.3 From 4111cdef0e871c0f7c8874b19ee68a3671f7d63e Mon Sep 17 00:00:00 2001 From: "Aneesh Kumar K.V" Date: Tue, 3 Sep 2019 18:04:51 +0530 Subject: powerpc/nvdimm: Use HCALL error as the return value This simplifies the error handling and also enable us to switch to H_SCM_QUERY hcall in a later patch on H_OVERLAP error. We also do some kernel print formatting fixup in this patch. Signed-off-by: Aneesh Kumar K.V Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20190903123452.28620-1-aneesh.kumar@linux.ibm.com --- arch/powerpc/platforms/pseries/papr_scm.c | 26 +++++++++++--------------- 1 file changed, 11 insertions(+), 15 deletions(-) (limited to 'arch/powerpc/platforms') diff --git a/arch/powerpc/platforms/pseries/papr_scm.c b/arch/powerpc/platforms/pseries/papr_scm.c index a5ac371a3f06..3bef4d298ac6 100644 --- a/arch/powerpc/platforms/pseries/papr_scm.c +++ b/arch/powerpc/platforms/pseries/papr_scm.c @@ -66,28 +66,22 @@ static int drc_pmem_bind(struct papr_scm_priv *p) } while (rc == H_BUSY); if (rc) { - /* H_OVERLAP needs a separate error path */ - if (rc == H_OVERLAP) - return -EBUSY; - dev_err(&p->pdev->dev, "bind err: %lld\n", rc); - return -ENXIO; + return rc; } p->bound_addr = saved; - - dev_dbg(&p->pdev->dev, "bound drc %x to %pR\n", p->drc_index, &p->res); - - return 0; + dev_dbg(&p->pdev->dev, "bound drc 0x%x to %pR\n", p->drc_index, &p->res); + return rc; } -static int drc_pmem_unbind(struct papr_scm_priv *p) +static void drc_pmem_unbind(struct papr_scm_priv *p) { unsigned long ret[PLPAR_HCALL_BUFSIZE]; uint64_t token = 0; int64_t rc; - dev_dbg(&p->pdev->dev, "unbind drc %x\n", p->drc_index); + dev_dbg(&p->pdev->dev, "unbind drc 0x%x\n", p->drc_index); /* NB: unbind has the same retry requirements as drc_pmem_bind() */ do { @@ -110,10 +104,10 @@ static int drc_pmem_unbind(struct papr_scm_priv *p) if (rc) dev_err(&p->pdev->dev, "unbind error: %lld\n", rc); else - dev_dbg(&p->pdev->dev, "unbind drc %x complete\n", + dev_dbg(&p->pdev->dev, "unbind drc 0x%x complete\n", p->drc_index); - return rc == H_SUCCESS ? 0 : -ENXIO; + return; } static int papr_scm_meta_get(struct papr_scm_priv *p, @@ -436,14 +430,16 @@ static int papr_scm_probe(struct platform_device *pdev) rc = drc_pmem_bind(p); /* If phyp says drc memory still bound then force unbound and retry */ - if (rc == -EBUSY) { + if (rc == H_OVERLAP) { dev_warn(&pdev->dev, "Retrying bind after unbinding\n"); drc_pmem_unbind(p); rc = drc_pmem_bind(p); } - if (rc) + if (rc != H_SUCCESS) { + rc = -ENXIO; goto err; + } /* setup the resource for the newly bound range */ p->res.start = p->bound_addr; -- cgit v1.2.3 From faa6d21153fd11e139dd880044521389b34a24f2 Mon Sep 17 00:00:00 2001 From: "Aneesh Kumar K.V" Date: Tue, 3 Sep 2019 18:04:52 +0530 Subject: powerpc/nvdimm: use H_SCM_QUERY hcall on H_OVERLAP error Right now we force an unbind of SCM memory at drcindex on H_OVERLAP error. This really slows down operations like kexec where we get the H_OVERLAP error because we don't go through a full hypervisor re init. H_OVERLAP error for a H_SCM_BIND_MEM hcall indicates that SCM memory at drc index is already bound. Since we don't specify a logical memory address for bind hcall, we can use the H_SCM_QUERY hcall to query the already bound logical address. Boot time difference with and without patch is: [ 5.583617] IOMMU table initialized, virtual merging enabled [ 5.603041] papr_scm ibm,persistent-memory:ibm,pmemory@44104001: Retrying bind after unbinding [ 301.514221] papr_scm ibm,persistent-memory:ibm,pmemory@44108001: Retrying bind after unbinding [ 340.057238] hv-24x7: read 1530 catalog entries, created 537 event attrs (0 failures), 275 descs after fix [ 5.101572] IOMMU table initialized, virtual merging enabled [ 5.116984] papr_scm ibm,persistent-memory:ibm,pmemory@44104001: Querying SCM details [ 5.117223] papr_scm ibm,persistent-memory:ibm,pmemory@44108001: Querying SCM details [ 5.120530] hv-24x7: read 1530 catalog entries, created 537 event attrs (0 failures), 275 descs Signed-off-by: Aneesh Kumar K.V Signed-off-by: Michael Ellerman Link: https://lore.kernel.org/r/20190903123452.28620-2-aneesh.kumar@linux.ibm.com --- arch/powerpc/platforms/pseries/papr_scm.c | 48 +++++++++++++++++++++++++------ 1 file changed, 40 insertions(+), 8 deletions(-) (limited to 'arch/powerpc/platforms') diff --git a/arch/powerpc/platforms/pseries/papr_scm.c b/arch/powerpc/platforms/pseries/papr_scm.c index 3bef4d298ac6..61883291defc 100644 --- a/arch/powerpc/platforms/pseries/papr_scm.c +++ b/arch/powerpc/platforms/pseries/papr_scm.c @@ -65,10 +65,8 @@ static int drc_pmem_bind(struct papr_scm_priv *p) cond_resched(); } while (rc == H_BUSY); - if (rc) { - dev_err(&p->pdev->dev, "bind err: %lld\n", rc); + if (rc) return rc; - } p->bound_addr = saved; dev_dbg(&p->pdev->dev, "bound drc 0x%x to %pR\n", p->drc_index, &p->res); @@ -110,6 +108,42 @@ static void drc_pmem_unbind(struct papr_scm_priv *p) return; } +static int drc_pmem_query_n_bind(struct papr_scm_priv *p) +{ + unsigned long start_addr; + unsigned long end_addr; + unsigned long ret[PLPAR_HCALL_BUFSIZE]; + int64_t rc; + + + rc = plpar_hcall(H_SCM_QUERY_BLOCK_MEM_BINDING, ret, + p->drc_index, 0); + if (rc) + goto err_out; + start_addr = ret[0]; + + /* Make sure the full region is bound. */ + rc = plpar_hcall(H_SCM_QUERY_BLOCK_MEM_BINDING, ret, + p->drc_index, p->blocks - 1); + if (rc) + goto err_out; + end_addr = ret[0]; + + if ((end_addr - start_addr) != ((p->blocks - 1) * p->block_size)) + goto err_out; + + p->bound_addr = start_addr; + dev_dbg(&p->pdev->dev, "bound drc 0x%x to %pR\n", p->drc_index, &p->res); + return rc; + +err_out: + dev_info(&p->pdev->dev, + "Failed to query, trying an unbind followed by bind"); + drc_pmem_unbind(p); + return drc_pmem_bind(p); +} + + static int papr_scm_meta_get(struct papr_scm_priv *p, struct nd_cmd_get_config_data_hdr *hdr) { @@ -430,13 +464,11 @@ static int papr_scm_probe(struct platform_device *pdev) rc = drc_pmem_bind(p); /* If phyp says drc memory still bound then force unbound and retry */ - if (rc == H_OVERLAP) { - dev_warn(&pdev->dev, "Retrying bind after unbinding\n"); - drc_pmem_unbind(p); - rc = drc_pmem_bind(p); - } + if (rc == H_OVERLAP) + rc = drc_pmem_query_n_bind(p); if (rc != H_SUCCESS) { + dev_err(&p->pdev->dev, "bind err: %d\n", rc); rc = -ENXIO; goto err; } -- cgit v1.2.3