Skip to contentSkip to conversation

powerpc/64: machine check and system reset fixes

PATCH Superseded Nicholas Piggin 3 participants 19 messages Mar 25, 2020 – Apr 3, 2020 PowerPC

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Nicholas Piggin

author reviewer
#0 lore raw

v2 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
author reviewer
#1 lore raw

PATCH 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
index 6a936c9199d6..67cbcb2d0c7f 100644
@@ -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
author reviewer
#3 lore raw

PATCH v2 2/12 powerpc/64s/exceptions: Change irq reconcile for NMIs from reusing _DAR to RESULT

In reply to Nicholas Piggin

A 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
index 67cbcb2d0c7f..257d24437130 100644
@@ -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
author reviewer
#4 lore raw

PATCH v2 3/12 powerpc/64s/exceptions: machine check reconcile irq state

In reply to Nicholas Piggin

pseries 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
index 257d24437130..687f1389a5e5 100644
@@ -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
author reviewer
#5 lore raw

PATCH v2 4/12 powerpc/pseries/ras: avoid calling rtas_token in NMI paths

In reply to 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.

---
 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
index ca33f4ef6cb4..6003c2e533a0 100644
@@ -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
index 1d7f973c647b..c74d5e740922 100644
@@ -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
index 17d17f064a2d..c31acd7ce0c0 100644
@@ -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
author reviewer
#7 lore raw

PATCH v2 5/12 powerpc/pseries/ras: FWNMI_VALID off by one

In reply to 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.

---
 arch/powerpc/platforms/pseries/ras.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/powerpc/platforms/pseries/ras.c +3 -2
index c74d5e740922..9a37bda47468 100644
@@ -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
author reviewer
#9 lore raw

PATCH v2 6/12 powerpc/pseries/ras: fwnmi avoid modifying r3 in error case

In reply to Nicholas Piggin

If 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
index 9a37bda47468..a40598e6e525 100644
@@ -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
author reviewer
#10 lore raw

PATCH v2 7/12 powerpc/pseries/ras: fwnmi sreset should not interlock

In reply to Nicholas Piggin

PAPR 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
index a40598e6e525..833ae34b7fec 100644
@@ -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
author reviewer
#11 lore raw

PATCH v2 8/12 powerpc/pseries: limit machine check stack to 4GB

In reply to Nicholas Piggin

This 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
index 3bf03666ee09..ca1041f8a578 100644
@@ -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
author reviewer
#14 lore raw

PATCH v2 9/12 powerpc/pseries: machine check use rtas_call_unlocked with args on stack

In reply to Nicholas Piggin

With 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
index 833ae34b7fec..6938261e4569 100644
@@ -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
author reviewer
#15 lore raw

PATCH v2 10/12 powerpc/64s: machine check interrupt update NMI accounting

In reply to Nicholas Piggin

machine_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
index 34c1001e9e8b..c1684be0d8b7 100644
@@ -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
index d27bf367e929..c21344c1a151 100644
@@ -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
index 82a3438300fd..1845fd7e161a 100644
@@ -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
author reviewer
#16 lore raw

PATCH v2 11/12 powerpc/64s: machine check do not trace real-mode handler

In reply to Nicholas Piggin

Rather 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
index c1684be0d8b7..36e99adb8710 100644
@@ -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
author reviewer
#17 lore raw

PATCH v2 12/12 powerpc/64s: system reset do not trace

In reply to 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.

---
 arch/powerpc/kernel/traps.c | 5 +++++
 1 file changed, 5 insertions(+)
arch/powerpc/kernel/traps.c +5 -0
index 1845fd7e161a..ed7b7a6e2dc0 100644
@@ -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
Acked-by on v2 12/12
#18 lore raw

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.

---
 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
index 1845fd7e161a..ed7b7a6e2dc0 100644
@@ -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

Acked-by on v2 1/12
#2 lore raw

Re: [PATCH v2 01/12] powerpc/64s/exceptions: Fix in_mce accounting in unrecoverable path

In reply to Nicholas Piggin
Quoted 10 lines · Nicholas Piggin
---
 arch/powerpc/kernel/exceptions-64s.S | 4 ++++
 1 file changed, 4 insertions(+)
arch/powerpc/kernel/exceptions-64s.S +0 -0
index 6a936c9199d6..67cbcb2d0c7f 100644

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.

Reviewed-by on v2 4/12
#6 lore raw

Re: [PATCH v2 04/12] powerpc/pseries/ras: avoid calling rtas_token in NMI paths

In reply to Nicholas Piggin
Quoted 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.

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
index ca33f4ef6cb4..6003c2e533a0 100644
@@ -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
index 1d7f973c647b..c74d5e740922 100644
@@ -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
index 17d17f064a2d..c31acd7ce0c0 100644
@@ -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
Acked-by on v2 5/12
#8 lore raw

Re: [PATCH v2 05/12] powerpc/pseries/ras: FWNMI_VALID off by one

In reply to Nicholas Piggin
Quoted 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.

---
 arch/powerpc/platforms/pseries/ras.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
arch/powerpc/platforms/pseries/ras.c +0 -0
index c74d5e740922..9a37bda47468 100644

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
Reviewed-by on v2 8/12
#12 lore raw

Re: [PATCH v2 08/12] powerpc/pseries: limit machine check stack to 4GB

In reply to Nicholas Piggin
Quoted 26 lines · Nicholas Piggin

This 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 +3 -0
index 3bf03666ee09..ca1041f8a578 100644
@@ -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.

author reviewer
#13 lore raw

Re: [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_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 +3 -0
index 3bf03666ee09..ca1041f8a578 100644
@@ -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