powerpc/64: machine check and system reset fixes
Nicholas Piggin
reviewerv2 cover powerpc/64: machine check and system reset fixes
There's a bunch of problems we hit bringing up fwnmi sreset and testing with mce injection on QEMU. Mostly pseries issues.
This series of fixes applies on top of next-test, the machine check reconcile patch won't apply cleanly to previous kernels but it might want to be backported. We can do that after upstreaming.
This doesn't solve all known problems yet, but fwnmi machine check and system reset injection in QEMU is significantly better. There will be more to come but these should be ready for review now.
Thanks,
Nick
Nicholas Piggin (12):
powerpc/64s/exceptions: Fix in_mce accounting in unrecoverable path powerpc/64s/exceptions: Change irq reconcile for NMIs from reusing
_DAR to RESULT
powerpc/64s/exceptions: machine check reconcile irq state powerpc/pseries/ras: avoid calling rtas_token in NMI paths powerpc/pseries/ras: FWNMI_VALID off by one powerpc/pseries/ras: fwnmi avoid modifying r3 in error case powerpc/pseries/ras: fwnmi sreset should not interlock powerpc/pseries: limit machine check stack to 4GB powerpc/pseries: machine check use rtas_call_unlocked with args on
stack
powerpc/64s: machine check interrupt update NMI accounting powerpc/64s: machine check do not trace real-mode handler powerpc/64s: system reset do not trace
arch/powerpc/include/asm/firmware.h | 1 + arch/powerpc/kernel/exceptions-64s.S | 33 +++++++++++--- arch/powerpc/kernel/mce.c | 13 +++++- arch/powerpc/kernel/process.c | 2 +- arch/powerpc/kernel/setup_64.c | 15 ++++++- arch/powerpc/kernel/traps.c | 18 +++----- arch/powerpc/platforms/pseries/ras.c | 62 +++++++++++++++++++------- arch/powerpc/platforms/pseries/setup.c | 13 ++++-- 8 files changed, 118 insertions(+), 39 deletions(-)
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 1/12 powerpc/64s/exceptions: Fix in_mce accounting in unrecoverable path
--- arch/powerpc/kernel/exceptions-64s.S | 4 ++++ 1 file changed, 4 insertions(+)
arch/powerpc/kernel/exceptions-64s.S +4 -0
@@ -1335,6 +1335,10 @@ END_FTR_SECTION_IFSET(CPU_FTR_HVMODE) andc r10,r10,r3 mtmsrd r10 + lhz r12,PACA_IN_MCE(r13) + subi r12,r12,1 + sth r12,PACA_IN_MCE(r13) + /* Invoke machine_check_exception to print MCE event and panic. */ addi r3,r1,STACK_FRAME_OVERHEAD bl machine_check_exception
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 2/12 powerpc/64s/exceptions: Change irq reconcile for NMIs from reusing _DAR to RESULT
In reply to Nicholas PigginA spare interrupt stack slot is needed to save irq state when reconciling NMIs (sreset and decrementer soft-nmi). _DAR is used for this, but we want to reconcile machine checks as well, which do use _DAR. Switch to using RESULT instead, as it's used by system calls.
--- arch/powerpc/kernel/exceptions-64s.S | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
arch/powerpc/kernel/exceptions-64s.S +5 -5
@@ -1011,13 +1011,13 @@ EXC_COMMON_BEGIN(system_reset_common) * the right thing. We do not want to reconcile because that goes * through irq tracing which we don't want in NMI. * - * Save PACAIRQHAPPENED to _DAR (otherwise unused), and set HARD_DIS + * Save PACAIRQHAPPENED to RESULT (otherwise unused), and set HARD_DIS * as we are running with MSR[EE]=0. */ li r10,IRQS_ALL_DISABLED stb r10,PACAIRQSOFTMASK(r13) lbz r10,PACAIRQHAPPENED(r13) - std r10,_DAR(r1) + std r10,RESULT(r1) ori r10,r10,PACA_IRQ_HARD_DIS stb r10,PACAIRQHAPPENED(r13)
@@ -1038,7 +1038,7 @@ EXC_COMMON_BEGIN(system_reset_common) /* * Restore soft mask settings. */ - ld r10,_DAR(r1) + ld r10,RESULT(r1) stb r10,PACAIRQHAPPENED(r13) ld r10,SOFTE(r1) stb r10,PACAIRQSOFTMASK(r13)
@@ -2809,7 +2809,7 @@ EXC_COMMON_BEGIN(soft_nmi_common) li r10,IRQS_ALL_DISABLED stb r10,PACAIRQSOFTMASK(r13) lbz r10,PACAIRQHAPPENED(r13) - std r10,_DAR(r1) + std r10,RESULT(r1) ori r10,r10,PACA_IRQ_HARD_DIS stb r10,PACAIRQHAPPENED(r13)
@@ -2823,7 +2823,7 @@ EXC_COMMON_BEGIN(soft_nmi_common) /* * Restore soft mask settings. */ - ld r10,_DAR(r1) + ld r10,RESULT(r1) stb r10,PACAIRQHAPPENED(r13) ld r10,SOFTE(r1) stb r10,PACAIRQSOFTMASK(r13)
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 3/12 powerpc/64s/exceptions: machine check reconcile irq state
In reply to Nicholas Pigginpseries fwnmi machine check code pops the soft-irq checks in rtas_call (after the previous patch to remove rtas_token from this call path). Rather than play whack a mole with these and forever having fragile code, it seems better to have the early machine check handler perform the same kind of reconcile as the other NMI interrupts.
WARNING: CPU: 0 PID: 493 at arch/powerpc/kernel/irq.c:343
CPU: 0 PID: 493 Comm: a Tainted: G W
NIP: c00000000001ed2c LR: c000000000042c40 CTR: 0000000000000000
REGS: c0000001fffd38b0 TRAP: 0700 Tainted: G W
MSR: 8000000000021003 <SF,ME,RI,LE> CR: 28000488 XER: 00000000
CFAR: c00000000001ec90 IRQMASK: 0
GPR00: c000000000043820 c0000001fffd3b40 c0000000012ba300 0000000000000000
GPR04: 0000000048000488 0000000000000000 0000000000000000 00000000deadbeef
GPR08: 0000000000000080 0000000000000000 0000000000000000 0000000000001001
GPR12: 0000000000000000 c0000000014a0000 0000000000000000 0000000000000000
GPR16: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
GPR20: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
GPR24: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
GPR28: 0000000000000000 0000000000000001 c000000001360810 0000000000000000
NIP [c00000000001ed2c] arch_local_irq_restore.part.0+0xac/0x100
LR [c000000000042c40] unlock_rtas+0x30/0x90
Call Trace:
[c0000001fffd3b40] [c000000001360810] 0xc000000001360810 (unreliable)
[c0000001fffd3b60] [c000000000043820] rtas_call+0x1c0/0x280
[c0000001fffd3bb0] [c0000000000dc328] fwnmi_release_errinfo+0x38/0x70
[c0000001fffd3c10] [c0000000000dcd8c] pseries_machine_check_realmode+0x1dc/0x540
[c0000001fffd3cd0] [c00000000003fe04] machine_check_early+0x54/0x70
[c0000001fffd3d00] [c000000000008384] machine_check_early_common+0x134/0x1f0
--- interrupt: 200 at 0x13f1307c8
LR = 0x7fff888b8528
Instruction dump:
60000000 7d2000a6 71298000 41820068 39200002 7d210164 4bffff9c 60000000
60000000 7d2000a6 71298000 4c820020 <0fe00000> 4e800020 60000000 60000000--- arch/powerpc/kernel/exceptions-64s.S | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
arch/powerpc/kernel/exceptions-64s.S +19 -0
@@ -1186,11 +1186,30 @@ END_FTR_SECTION_IFSET(CPU_FTR_HVMODE) li r10,MSR_RI mtmsrd r10,1 + /* + * Set IRQS_ALL_DISABLED and save PACAIRQHAPPENED (see + * system_reset_common) + */ + li r10,IRQS_ALL_DISABLED + stb r10,PACAIRQSOFTMASK(r13) + lbz r10,PACAIRQHAPPENED(r13) + std r10,RESULT(r1) + ori r10,r10,PACA_IRQ_HARD_DIS + stb r10,PACAIRQHAPPENED(r13) + addi r3,r1,STACK_FRAME_OVERHEAD bl machine_check_early std r3,RESULT(r1) /* Save result */ ld r12,_MSR(r1) + /* + * Restore soft mask settings. + */ + ld r10,RESULT(r1) + stb r10,PACAIRQHAPPENED(r13) + ld r10,SOFTE(r1) + stb r10,PACAIRQSOFTMASK(r13) + #ifdef CONFIG_PPC_P7_NAP /* * Check if thread was in power saving mode. We come here when any
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 4/12 powerpc/pseries/ras: avoid calling rtas_token in NMI paths
In reply to Nicholas PigginIn the interest of reducing code and possible failures in the machine check and system reset paths, grab the "ibm,nmi-interlock" token at init time.
--- arch/powerpc/include/asm/firmware.h | 1 + arch/powerpc/platforms/pseries/ras.c | 2 +- arch/powerpc/platforms/pseries/setup.c | 13 ++++++++++--- 3 files changed, 12 insertions(+), 4 deletions(-)
arch/powerpc/include/asm/firmware.h +1 -0
@@ -128,6 +128,7 @@ extern void machine_check_fwnmi(void);/* This is true if we are using the firmware NMI handler (typically LPAR) */ extern int fwnmi_active; +extern int ibm_nmi_interlock_token;extern unsigned int __start___fw_ftr_fixup, __stop___fw_ftr_fixup;
arch/powerpc/platforms/pseries/ras.c +1 -1
@@ -458,7 +458,7 @@ static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)*/ static void fwnmi_release_errinfo (void ) { -int ret =rtas_call (rtas_token ("ibm,nmi-interlock" ),0 ,1 ,NULL ); +int ret =rtas_call (ibm_nmi_interlock_token,0 ,1 ,NULL );if (ret !=0 )printk (KERN_ERR "FWNMI: nmi-interlock failed: %d\n" , ret); }
arch/powerpc/platforms/pseries/setup.c +10 -3
@@ -83,6 +83,7 @@ unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);EXPORT_SYMBOL (CMO_PageSize);int fwnmi_active;/* TRUE if an FWNMI handler is present */ +int ibm_nmi_interlock_token;static void pSeries_show_cpuinfo (struct seq_file *m) {
@@ -113,9 +114,14 @@ static void __init fwnmi_init(void)struct slb_entry *slb_ptr;size_t size;#endif +int ibm_nmi_register_token; -int ibm_nmi_register =rtas_token ("ibm,nmi-register" ); -if (ibm_nmi_register ==RTAS_UNKNOWN_SERVICE ) + ibm_nmi_register_token =rtas_token ("ibm,nmi-register" ); +if (ibm_nmi_register_token ==RTAS_UNKNOWN_SERVICE ) +return ; + + ibm_nmi_interlock_token =rtas_token ("ibm,nmi-interlock" ); +if (WARN_ON (ibm_nmi_interlock_token ==RTAS_UNKNOWN_SERVICE ))return ;/* If the kernel's not linked at zero we point the firmware at low
@@ -123,7 +129,8 @@ static void __init fwnmi_init(void) system_reset_addr =__pa (system_reset_fwnmi) -PHYSICAL_START ; machine_check_addr =__pa (machine_check_fwnmi) -PHYSICAL_START ; -if (0 ==rtas_call (ibm_nmi_register,2 ,1 ,NULL , system_reset_addr, +if (0 ==rtas_call (ibm_nmi_register_token,2 ,1 ,NULL , + system_reset_addr, machine_check_addr)) fwnmi_active =1 ;
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 5/12 powerpc/pseries/ras: FWNMI_VALID off by one
In reply to Nicholas PigginThis was discovered developing qemu fwnmi sreset support. This off-by-one bug means the last 16 bytes of the rtas area can not be used for a 16 byte save area.
It's not a serious bug, and QEMU implementation has to retain a workaround for old kernels, but it's good to tighten it.
--- arch/powerpc/platforms/pseries/ras.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
arch/powerpc/platforms/pseries/ras.c +3 -2
@@ -395,10 +395,11 @@ static irqreturn_t ras_error_interrupt(int irq, void *dev_id)/* * Some versions of FWNMI place the buffer inside the 4kB page starting at * 0x7000. Other versions place it inside the rtas buffer. We check both. +* Minimum size of the buffer is 16 bytes. */ #define VALID_FWNMI_BUFFER (A ) \ - ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \ - (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16)))) + ((((A) >= 0x7000) && ((A) <= 0x8000 - 16)) || \ + (((A) >= rtas.base) && ((A) <= (rtas.base + rtas.size - 16))))static inline struct rtas_error_log *fwnmi_get_errlog (void ) {
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 6/12 powerpc/pseries/ras: fwnmi avoid modifying r3 in error case
In reply to Nicholas PigginIf there is some error with the fwnmi save area, r3 has already been modified which doesn't help with debugging.
Only update r3 when to restore the saved value.
--- arch/powerpc/platforms/pseries/ras.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
arch/powerpc/platforms/pseries/ras.c +4 -3
@@ -423,18 +423,19 @@ static inline struct rtas_error_log *fwnmi_get_errlog(void)*/ static struct rtas_error_log *fwnmi_get_errinfo (struct pt_regs *regs) { +unsigned long savep_ra;unsigned long *savep;struct rtas_error_log *h;/* Mask top two bits */ - regs->gpr[3 ] &= ~(0x3UL <<62 ); + savep_ra = regs->gpr[3 ] & ~(0x3UL <<62 ); -if (!VALID_FWNMI_BUFFER (regs->gpr[3 ])) { +if (!VALID_FWNMI_BUFFER (savep_ra)) {printk (KERN_ERR "FWNMI: corrupt r3 0x%016lx\n" , regs->gpr[3 ]);return NULL ; } - savep =__va (regs->gpr[3 ]); + savep =__va (savep_ra); regs->gpr[3 ] =be64_to_cpu (savep[0 ]);/* restore original r3 */ h = (struct rtas_error_log *)&savep[1 ];
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 7/12 powerpc/pseries/ras: fwnmi sreset should not interlock
In reply to Nicholas PigginPAPR does not specify that fwnmi sreset should be interlocked, and PowerVM (and therefore now QEMU) do not require it.
These "ibm,nmi-interlock" calls are ignored by firmware, but there is a possibility that the sreset could have interrupted a machine check and release the machine check's interlock too early, corrupting it if another machine check came in.
This is an extremely rare case, but it should be fixed for clarity and reducing the code executed in the sreset path. Firmware also does not provide error information for the sreset case to look at, so remove that comment.
--- arch/powerpc/platforms/pseries/ras.c | 48 ++++++++++++++++++++-------- 1 file changed, 34 insertions(+), 14 deletions(-)
arch/powerpc/platforms/pseries/ras.c +34 -14
@@ -406,6 +406,20 @@ static inline struct rtas_error_log *fwnmi_get_errlog(void)return (struct rtas_error_log *)local_paca->mce_data_buf; } +static unsigned long *fwnmi_get_savep (struct pt_regs *regs) +{ +unsigned long savep_ra; + +/* Mask top two bits */ + savep_ra = regs->gpr[3 ] & ~(0x3UL <<62 ); +if (!VALID_FWNMI_BUFFER (savep_ra)) { +printk (KERN_ERR "FWNMI: corrupt r3 0x%016lx\n" , regs->gpr[3 ]); +return NULL ; + } + +return __va (savep_ra); +} +/* * Get the error information for errors coming through the * FWNMI vectors. The pt_regs' r3 will be updated to reflect
@@ -423,20 +437,15 @@ static inline struct rtas_error_log *fwnmi_get_errlog(void)*/ static struct rtas_error_log *fwnmi_get_errinfo (struct pt_regs *regs) { -unsigned long savep_ra;unsigned long *savep;struct rtas_error_log *h; -/* Mask top two bits */ - savep_ra = regs->gpr[3 ] & ~(0x3UL <<62 ); - -if (!VALID_FWNMI_BUFFER (savep_ra)) { -printk (KERN_ERR "FWNMI: corrupt r3 0x%016lx\n" , regs->gpr[3 ]); + savep =fwnmi_get_savep (regs); +if (!savep)return NULL ; - } - savep =__va (savep_ra); - regs->gpr[3 ] =be64_to_cpu (savep[0 ]);/* restore original r3 */ +/* restore original r3 */ + regs->gpr[3 ] =be64_to_cpu (savep[0 ]); h = (struct rtas_error_log *)&savep[1 ];/* Use the per cpu buffer from paca to store rtas error log */
@@ -483,11 +492,22 @@ int pSeries_system_reset_exception(struct pt_regs *regs)#endif if (fwnmi_active) { -struct rtas_error_log *errhdr =fwnmi_get_errinfo (regs); -if (errhdr) { -/* XXX Should look at FWNMI information */ - } -fwnmi_release_errinfo (); +unsigned long *savep; + +/* +* Firmware (PowerVM and KVM) saves r3 to a save area like +* machine check, which is not exactly what PAPR (2.9) +* suggests but there is no way to detect otherwise, so this +* is the interface now. +* +* System resets do not save any error log or require an +* "ibm,nmi-interlock" rtas call to release. +*/ + + savep =fwnmi_get_savep (regs); +/* restore original r3 */ +if (savep) + regs->gpr[3 ] =be64_to_cpu (savep[0 ]); }if (smp_handle_nmi_ipi (regs))
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 8/12 powerpc/pseries: limit machine check stack to 4GB
In reply to Nicholas PigginThis allows rtas_args to be put on the machine check stack, which avoids a lot of complications with re-entrancy deadlocks.
--- arch/powerpc/kernel/setup_64.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-)
arch/powerpc/kernel/setup_64.c +14 -1
@@ -695,6 +695,9 @@ void __init exc_lvl_early_init(void)void __init emergency_stack_init (void ) {u64 limit; +#ifdef CONFIG_PPC_BOOK3S_64 +u64 mce_limit; +#endif unsigned int i;/*
@@ -713,6 +716,16 @@ void __init emergency_stack_init(void)*/ limit =min (ppc64_bolted_size (), ppc64_rma_size); +/* +* Machine check on pseries calls rtas, but can't use the static +* rtas_args due to a machine check hitting while the lock is held. +* rtas args have to be under 4GB, so the machine check stack is +* limited to 4GB so args can be put on stack. +*/ + mce_limit = limit; +if (firmware_has_feature (FW_FEATURE_LPAR ) && mce_limit >4UL *1024 *1024 *1024 ) + mce_limit =4UL *1024 *1024 *1024 ; +for_each_possible_cpu (i) { paca_ptrs[i]->emergency_sp =alloc_stack (limit, i) +THREAD_SIZE ;
@@ -721,7 +734,7 @@ void __init emergency_stack_init(void) paca_ptrs[i]->nmi_emergency_sp =alloc_stack (limit, i) +THREAD_SIZE ;/* emergency stack for machine check exception handling. */ - paca_ptrs[i]->mc_emergency_sp =alloc_stack (limit, i) +THREAD_SIZE ; + paca_ptrs[i]->mc_emergency_sp =alloc_stack (mce_limit, i) +THREAD_SIZE ;#endif } }
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 9/12 powerpc/pseries: machine check use rtas_call_unlocked with args on stack
In reply to Nicholas PigginWith the previous patch, machine checks can use rtas_call_unlocked which avoids the rtas spinlock which would deadlock if a machine check hits while making an rtas call.
This also avoids the complex rtas error logging which has more rtas calls and includes kmalloc (which can return memory beyond RMA, which would also crash).
--- arch/powerpc/platforms/pseries/ras.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-)
arch/powerpc/platforms/pseries/ras.c +9 -1
@@ -469,7 +469,15 @@ static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)*/ static void fwnmi_release_errinfo (void ) { -int ret =rtas_call (ibm_nmi_interlock_token,0 ,1 ,NULL ); +struct rtas_args rtas_args; +int ret; + +/* +* On pseries, the machine check stack is limited to under 4GB, so +* args can be on-stack. +*/ +rtas_call_unlocked (&rtas_args, ibm_nmi_interlock_token,0 ,1 ,NULL ); + ret =be32_to_cpu (rtas_args.rets[0 ]);if (ret !=0 )printk (KERN_ERR "FWNMI: nmi-interlock failed: %d\n" , ret); }
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 10/12 powerpc/64s: machine check interrupt update NMI accounting
In reply to Nicholas Pigginmachine_check_early is taken as an NMI, so nmi_enter is used there. machine_check_exception is no longer taken as an NMI (it's invoked via irq_work in the case a machine check hits in kernel mode), so remove the nmi_enter from that case.
In NMI context, hash faults don't try to refill the hash table, which can lead to crashes accessing non-pinned kernel pages. System reset still has this potential problem.
--- arch/powerpc/kernel/mce.c | 7 +++++++ arch/powerpc/kernel/process.c | 2 +- arch/powerpc/kernel/traps.c | 13 +------------ 3 files changed, 9 insertions(+), 13 deletions(-)
arch/powerpc/kernel/mce.c +7 -0
@@ -560,6 +560,9 @@ EXPORT_SYMBOL_GPL(machine_check_print_event_info);long machine_check_early (struct pt_regs *regs) {long handled =0 ; +bool nested =in_nmi (); +if (!nested) +nmi_enter ();hv_nmi_check_nonrecoverable (regs);
@@ -568,6 +571,10 @@ long machine_check_early(struct pt_regs *regs)*/ if (ppc_md.machine_check_early) handled = ppc_md.machine_check_early (regs); + +if (!nested) +nmi_exit (); +return handled; }
arch/powerpc/kernel/process.c +1 -1
@@ -1421,7 +1421,7 @@ void show_regs(struct pt_regs * regs)pr_cont ("DAR: " REG " DSISR: %08lx " , regs->dar, regs->dsisr);#endif #ifdef CONFIG_PPC64 -pr_cont ("IRQMASK: %lx " , regs->softe); +pr_cont ("IRQMASK: %lx IN_NMI:%d IN_MCE:%d" , regs->softe, (int )get_paca ()->in_nmi, (int )get_paca ()->in_mce);#endif #ifdef CONFIG_PPC_TRANSACTIONAL_MEM if (MSR_TM_ACTIVE (regs->msr))
arch/powerpc/kernel/traps.c +1 -12
@@ -823,9 +823,6 @@ int machine_check_generic(struct pt_regs *regs)void machine_check_exception (struct pt_regs *regs) {int recover =0 ; -bool nested =in_nmi (); -if (!nested) -nmi_enter ();__this_cpu_inc (irq_stat.mce_exceptions);
@@ -851,20 +848,12 @@ void machine_check_exception(struct pt_regs *regs)if (check_io_access (regs)) goto bail; -if (!nested) -nmi_exit (); -die ("Machine check" , regs,SIGBUS ); +bail:/* Must die if the interrupt is not recoverable */ if (!(regs->msr &MSR_RI ))nmi_panic (regs,"Unrecoverable Machine check" ); - -return ; - -bail: -if (!nested) -nmi_exit (); }void SMIException (struct pt_regs *regs)
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 11/12 powerpc/64s: machine check do not trace real-mode handler
In reply to Nicholas PigginRather than notrace annotations throughout a significant part of the machine check code across kernel/ pseries/ and powernv/ which can easily be broken and is infrequently tested, use paca->ftrace_enabled to blanket-disable tracing of the real-mode non-maskable handler.
--- arch/powerpc/kernel/mce.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
arch/powerpc/kernel/mce.c +5 -1
@@ -557,10 +557,13 @@ EXPORT_SYMBOL_GPL(machine_check_print_event_info);* * regs->nip and regs->msr contains srr0 and ssr1. */ -long machine_check_early (struct pt_regs *regs) +long notrace machine_check_early (struct pt_regs *regs) {long handled =0 ;bool nested =in_nmi (); +u8 ftrace_enabled = local_paca->ftrace_enabled; + + local_paca->ftrace_enabled =0 ;if (!nested)nmi_enter ();
@@ -574,6 +577,7 @@ long machine_check_early(struct pt_regs *regs)if (!nested)nmi_exit (); + local_paca->ftrace_enabled = ftrace_enabled;return handled; }
Signature · 1 line
2.23.0
Nicholas Piggin
reviewerPATCH v2 12/12 powerpc/64s: system reset do not trace
In reply to Nicholas PigginSimilarly to the previous patch, do not trace system reset. This code is used when there is a crash or hang, and tracing disturbs the system more and has been known to crash in the crash handling path.
--- arch/powerpc/kernel/traps.c | 5 +++++ 1 file changed, 5 insertions(+)
arch/powerpc/kernel/traps.c +5 -0
@@ -443,6 +443,9 @@ void system_reset_exception(struct pt_regs *regs)unsigned long hsrr0, hsrr1;bool nested =in_nmi ();bool saved_hsrrs = false; +u8 ftrace_enabled = local_paca->ftrace_enabled; + + local_paca->ftrace_enabled =0 ;/* * Avoid crashes in case of nested NMI exceptions. Recoverability
@@ -524,6 +527,8 @@ void system_reset_exception(struct pt_regs *regs)if (!nested)nmi_exit (); + local_paca->ftrace_enabled = ftrace_enabled; +/* What should we do here? We could issue a shutdown or hard reset. */ }
Signature · 1 line
2.23.0
Re: [PATCH v2 12/12] powerpc/64s: system reset do not trace
Quoted 8 lines · Nicholas Piggin
Similarly to the previous patch, do not trace system reset. This code is used when there is a crash or hang, and tracing disturbs the system more and has been known to crash in the crash handling path.
Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/kernel/traps.c | 5 +++++ 1 file changed, 5 insertions(+)
For the ftrace bits:
Quoted 16 lines
arch/powerpc/kernel/traps.c +3
-0@@ -443,6 +443,9 @@ void system_reset_exception(struct pt_regs *regs)unsigned long hsrr0, hsrr1;bool nested =in_nmi ();bool saved_hsrrs = false; +u8 ftrace_enabled = local_paca->ftrace_enabled; + + local_paca->ftrace_enabled =0 ;/* * Avoid crashes in case of nested NMI exceptions. Recoverability
arch/powerpc/kernel/traps.c void system_reset_exception(struct pt_regs *regs) · line 531
Quoted 5 lines
arch/powerpc/kernel/traps.c +2
-0@@ -524,6 +527,8 @@ void system_reset_exception(struct pt_regs *regs)if (!nested)nmi_exit (); + local_paca->ftrace_enabled = ftrace_enabled; +
I suppose we could add helpers to save/restore the state. But, since we only have these two users as of now, we can revisit that if the need arises.
- Naveen
Re: [PATCH v2 01/12] powerpc/64s/exceptions: Fix in_mce accounting in unrecoverable path
In reply to Nicholas PigginQuoted 10 lines · Nicholas Piggin
Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/kernel/exceptions-64s.S | 4 ++++ 1 file changed, 4 insertions(+)arch/powerpc/kernel/exceptions-64s.S +0
-0
arch/powerpc/kernel/exceptions-64s.S END_FTR_SECTION_IFSET(CPU_FTR_HVMODE) · line 1341
Quoted 7 lines
arch/powerpc/kernel/exceptions-64s.S +4
-0@@ -1335,6 +1335,10 @@ END_FTR_SECTION_IFSET(CPU_FTR_HVMODE) andc r10,r10,r3 mtmsrd r10 + lhz r12,PACA_IN_MCE(r13) + subi r12,r12,1 + sth r12,PACA_IN_MCE(r13) +
Thanks,
-Mahesh.
Re: [PATCH v2 04/12] powerpc/pseries/ras: avoid calling rtas_token in NMI paths
In reply to Nicholas PigginQuoted 5 lines · Nicholas Piggin
In the interest of reducing code and possible failures in the machine check and system reset paths, grab the "ibm,nmi-interlock" token at init time.
Signed-off-by Nicholas Piggin <[email protected]>
Thanks,
-Mahesh.
Quoted 73 lines
--- arch/powerpc/include/asm/firmware.h | 1 + arch/powerpc/platforms/pseries/ras.c | 2 +- arch/powerpc/platforms/pseries/setup.c | 13 ++++++++++--- 3 files changed, 12 insertions(+), 4 deletions(-)arch/powerpc/include/asm/firmware.h +1
-0@@ -128,6 +128,7 @@ extern void machine_check_fwnmi(void);/* This is true if we are using the firmware NMI handler (typically LPAR) */ extern int fwnmi_active; +extern int ibm_nmi_interlock_token;extern unsigned int __start___fw_ftr_fixup, __stop___fw_ftr_fixup;arch/powerpc/platforms/pseries/ras.c +1
-1@@ -458,7 +458,7 @@ static struct rtas_error_log *fwnmi_get_errinfo(struct pt_regs *regs)*/ static void fwnmi_release_errinfo (void ) { -int ret =rtas_call (rtas_token ("ibm,nmi-interlock" ),0 ,1 ,NULL ); +int ret =rtas_call (ibm_nmi_interlock_token,0 ,1 ,NULL );if (ret !=0 )printk (KERN_ERR "FWNMI: nmi-interlock failed: %d\n" , ret); }arch/powerpc/platforms/pseries/setup.c +10
-3@@ -83,6 +83,7 @@ unsigned long CMO_PageSize = (ASM_CONST(1) << IOMMU_PAGE_SHIFT_4K);EXPORT_SYMBOL (CMO_PageSize);int fwnmi_active;/* TRUE if an FWNMI handler is present */ +int ibm_nmi_interlock_token;static void pSeries_show_cpuinfo (struct seq_file *m) {@@ -113,9 +114,14 @@ static void __init fwnmi_init(void)struct slb_entry *slb_ptr;size_t size;#endif +int ibm_nmi_register_token; -int ibm_nmi_register =rtas_token ("ibm,nmi-register" ); -if (ibm_nmi_register ==RTAS_UNKNOWN_SERVICE ) + ibm_nmi_register_token =rtas_token ("ibm,nmi-register" ); +if (ibm_nmi_register_token ==RTAS_UNKNOWN_SERVICE ) +return ; + + ibm_nmi_interlock_token =rtas_token ("ibm,nmi-interlock" ); +if (WARN_ON (ibm_nmi_interlock_token ==RTAS_UNKNOWN_SERVICE ))return ;/* If the kernel's not linked at zero we point the firmware at low @@ -123,7 +129,8 @@ static void __init fwnmi_init(void) system_reset_addr =__pa (system_reset_fwnmi) -PHYSICAL_START ; machine_check_addr =__pa (machine_check_fwnmi) -PHYSICAL_START ; -if (0 ==rtas_call (ibm_nmi_register,2 ,1 ,NULL , system_reset_addr, +if (0 ==rtas_call (ibm_nmi_register_token,2 ,1 ,NULL , + system_reset_addr, machine_check_addr)) fwnmi_active =1 ;Signature · 1 line
2.23.0
Signature · 1 line
Mahesh J Salgaonkar
Re: [PATCH v2 05/12] powerpc/pseries/ras: FWNMI_VALID off by one
In reply to Nicholas PigginQuoted 17 lines · Nicholas Piggin
This was discovered developing qemu fwnmi sreset support. This off-by-one bug means the last 16 bytes of the rtas area can not be used for a 16 byte save area.
It's not a serious bug, and QEMU implementation has to retain a workaround for old kernels, but it's good to tighten it.
Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/platforms/pseries/ras.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)arch/powerpc/platforms/pseries/ras.c +0
-0
arch/powerpc/platforms/pseries/ras.c static irqreturn_t ras_error_interrupt(int irq, void *dev_id) · line 398
Quoted 4 lines
arch/powerpc/platforms/pseries/ras.c +1
-0@@ -395,10 +395,11 @@ static irqreturn_t ras_error_interrupt(int irq, void *dev_id)/* * Some versions of FWNMI place the buffer inside the 4kB page starting at * 0x7000. Other versions place it inside the rtas buffer. We check both. +* Minimum size of the buffer is 16 bytes.
Thanks,
-Mahesh.
Quoted 12 lines
arch/powerpc/platforms/pseries/ras.c +2
-2*/ #define VALID_FWNMI_BUFFER (A ) \ - ((((A) >= 0x7000) && ((A) < 0x7ff0)) || \ - (((A) >= rtas.base) && ((A) < (rtas.base + rtas.size - 16)))) + ((((A) >= 0x7000) && ((A) <= 0x8000 - 16)) || \ + (((A) >= rtas.base) && ((A) <= (rtas.base + rtas.size - 16))))static inline struct rtas_error_log *fwnmi_get_errlog (void ) {Signature · 1 line
2.23.0
Signature · 1 line
Mahesh J Salgaonkar
Re: [PATCH v2 08/12] powerpc/pseries: limit machine check stack to 4GB
In reply to Nicholas PigginQuoted 26 lines · Nicholas Piggin
This allows
rtas_argsto be put on the machine check stack, which avoids a lot of complications with re-entrancy deadlocks.Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/kernel/setup_64.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-)arch/powerpc/kernel/setup_64.c +3
-0@@ -695,6 +695,9 @@ void __init exc_lvl_early_init(void)void __init emergency_stack_init (void ) {u64 limit; +#ifdef CONFIG_PPC_BOOK3S_64 +u64 mce_limit; +#endif unsigned int i;/* @@ -713,6 +716,16 @@ void __init emergency_stack_init(void)*/ limit =min (ppc64_bolted_size (), ppc64_rma_size);
arch/powerpc/kernel/setup_64.c void __init emergency_stack_init(void) · line 728
Quoted 10 lines
arch/powerpc/kernel/setup_64.c +10
-0+/* +* Machine check on pseries calls rtas, but can't use the static +* rtas_args due to a machine check hitting while the lock is held. +* rtas args have to be under 4GB, so the machine check stack is +* limited to 4GB so args can be put on stack. +*/ + mce_limit = limit; +if (firmware_has_feature (FW_FEATURE_LPAR ) && mce_limit >4UL *1024 *1024 *1024 ) + mce_limit =4UL *1024 *1024 *1024 ; +
Don't you need this as well under CONFIG_PPC_BOOK3S_64 #ifdef ??
Rest looks good.
Thanks,
-Mahesh.
Nicholas Piggin
reviewerRe: [PATCH v2 08/12] powerpc/pseries: limit machine check stack to 4GB
Mahesh J Salgaonkar's on March 27, 2020 3:24 pm:
Quoted 27 lines · Nicholas Piggin
Quoted 26 lines
This allows
rtas_argsto be put on the machine check stack, which avoids a lot of complications with re-entrancy deadlocks.Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/kernel/setup_64.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-)arch/powerpc/kernel/setup_64.c +3
-0@@ -695,6 +695,9 @@ void __init exc_lvl_early_init(void)void __init emergency_stack_init (void ) {u64 limit; +#ifdef CONFIG_PPC_BOOK3S_64 +u64 mce_limit; +#endif unsigned int i;/* @@ -713,6 +716,16 @@ void __init emergency_stack_init(void)*/ limit =min (ppc64_bolted_size (), ppc64_rma_size);
Quoted 12 lines
arch/powerpc/kernel/setup_64.c void __init emergency_stack_init(void) · line 728
Quoted 10 lines
arch/powerpc/kernel/setup_64.c +10
-0+/* +* Machine check on pseries calls rtas, but can't use the static +* rtas_args due to a machine check hitting while the lock is held. +* rtas args have to be under 4GB, so the machine check stack is +* limited to 4GB so args can be put on stack. +*/ + mce_limit = limit; +if (firmware_has_feature (FW_FEATURE_LPAR ) && mce_limit >4UL *1024 *1024 *1024 ) + mce_limit =4UL *1024 *1024 *1024 ; +Don't you need this as well under
CONFIG_PPC_BOOK3S_64#ifdef ??
Good catch.
Thanks,
Nick