[PATCH v2] powerpc/book3s: Recover from MC in opal on SCOM read via MMIO.
Benjamin Herrenschmidt
benh at kernel.crashing.org
Sat Mar 29 22:41:31 EST 2014
On Sat, 2014-03-29 at 10:27 +0530, Mahesh J Salgaonkar wrote:
> From: Mahesh Salgaonkar <mahesh at linux.vnet.ibm.com>
>
> Detect and recover from machine check when inside opal on a special
> scom load instructions. On specific SCOM read via MMIO we may get a machine
> check exception with SRR0 pointing inside opal. To recover from MC
> in this scenario, get a recovery instruction address and return to it from
> MC.
V1 is already in -next, please send an incremental patch.
Cheers,
Ben.
> OPAL will export the machine check recoverable ranges through
> device tree node mcheck-recoverable-ranges under ibm,opal:
>
> # hexdump /proc/device-tree/ibm,opal/mcheck-recoverable-ranges
> 0000000 0000 0000 3000 2804 0000 000c 0000 0000
> 0000010 3000 2814 0000 0000 3000 27f0 0000 000c
> 0000020 0000 0000 3000 2814 xxxx xxxx xxxx xxxx
> 0000030 llll llll yyyy yyyy yyyy yyyy
> ...
> ...
> #
>
> where:
> xxxx xxxx xxxx xxxx = Starting instruction address
> llll llll = Length of the address range.
> yyyy yyyy yyyy yyyy = recovery address
>
> Each recoverable address range entry is (start address, len,
> recovery address), 2 cells each for start and recovery address, 1 cell for
> len, totalling 5 cells per entry. During kernel boot time, build up the
> recovery table with the list of recovery ranges from device-tree node which
> will be used during machine check exception to recover from MMIO SCOM UE.
>
> Changes in V2:
> - Allocate mc_recoverable_range buffer based on number of entries
> of recoverable ranges instead of device property size. Without this
> change we end up allocating less memory and run into buffer corruption
> issue.
>
> Signed-off-by: Mahesh Salgaonkar <mahesh at linux.vnet.ibm.com>
> ---
> arch/powerpc/include/asm/machdep.h | 3 +
> arch/powerpc/include/asm/mce.h | 3 +
> arch/powerpc/include/asm/opal.h | 3 +
> arch/powerpc/kernel/mce.c | 4 +
> arch/powerpc/kernel/mce_power.c | 37 +++++++++--
> arch/powerpc/kernel/prom.c | 5 +
> arch/powerpc/platforms/powernv/opal.c | 112 +++++++++++++++++++++++++++++++-
> arch/powerpc/platforms/powernv/setup.c | 1
> 8 files changed, 158 insertions(+), 10 deletions(-)
>
> diff --git a/arch/powerpc/include/asm/machdep.h b/arch/powerpc/include/asm/machdep.h
> index ad3025d..4da6574 100644
> --- a/arch/powerpc/include/asm/machdep.h
> +++ b/arch/powerpc/include/asm/machdep.h
> @@ -170,6 +170,9 @@ struct machdep_calls {
> int (*system_reset_exception)(struct pt_regs *regs);
> int (*machine_check_exception)(struct pt_regs *regs);
>
> + /* Called during machine check exception to retrive fixup address. */
> + bool (*mce_check_early_recovery)(struct pt_regs *regs);
> +
> /* Motherboard/chipset features. This is a kind of general purpose
> * hook used to control some machine specific features (like reset
> * lines, chip power control, etc...).
> diff --git a/arch/powerpc/include/asm/mce.h b/arch/powerpc/include/asm/mce.h
> index 8e99edf..f97d8cb 100644
> --- a/arch/powerpc/include/asm/mce.h
> +++ b/arch/powerpc/include/asm/mce.h
> @@ -187,7 +187,8 @@ struct mce_error_info {
> #define MCE_EVENT_DONTRELEASE false
>
> extern void save_mce_event(struct pt_regs *regs, long handled,
> - struct mce_error_info *mce_err, uint64_t addr);
> + struct mce_error_info *mce_err, uint64_t nip,
> + uint64_t addr);
> extern int get_mce_event(struct machine_check_event *mce, bool release);
> extern void release_mce_event(void);
> extern void machine_check_queue_event(void);
> diff --git a/arch/powerpc/include/asm/opal.h b/arch/powerpc/include/asm/opal.h
> index ed82142..ad67c40 100644
> --- a/arch/powerpc/include/asm/opal.h
> +++ b/arch/powerpc/include/asm/opal.h
> @@ -833,6 +833,8 @@ int64_t opal_sync_host_reboot(void);
>
> /* Internal functions */
> extern int early_init_dt_scan_opal(unsigned long node, const char *uname, int depth, void *data);
> +extern int early_init_dt_scan_recoverable_ranges(unsigned long node,
> + const char *uname, int depth, void *data);
>
> extern int opal_get_chars(uint32_t vtermno, char *buf, int count);
> extern int opal_put_chars(uint32_t vtermno, const char *buf, int total_len);
> @@ -863,6 +865,7 @@ extern void opal_nvram_init(void);
> extern void opal_flash_init(void);
>
> extern int opal_machine_check(struct pt_regs *regs);
> +extern bool opal_mce_check_early_recovery(struct pt_regs *regs);
>
> extern void opal_shutdown(void);
>
> diff --git a/arch/powerpc/kernel/mce.c b/arch/powerpc/kernel/mce.c
> index cadef7e..a7fd4cb 100644
> --- a/arch/powerpc/kernel/mce.c
> +++ b/arch/powerpc/kernel/mce.c
> @@ -70,7 +70,7 @@ static void mce_set_error_info(struct machine_check_event *mce,
> */
> void save_mce_event(struct pt_regs *regs, long handled,
> struct mce_error_info *mce_err,
> - uint64_t addr)
> + uint64_t nip, uint64_t addr)
> {
> uint64_t srr1;
> int index = __get_cpu_var(mce_nest_count)++;
> @@ -86,7 +86,7 @@ void save_mce_event(struct pt_regs *regs, long handled,
>
> /* Populate generic machine check info */
> mce->version = MCE_V1;
> - mce->srr0 = regs->nip;
> + mce->srr0 = nip;
> mce->srr1 = regs->msr;
> mce->gpr3 = regs->gpr[3];
> mce->in_use = 1;
> diff --git a/arch/powerpc/kernel/mce_power.c b/arch/powerpc/kernel/mce_power.c
> index 27c93f4..aa9aff3 100644
> --- a/arch/powerpc/kernel/mce_power.c
> +++ b/arch/powerpc/kernel/mce_power.c
> @@ -26,6 +26,7 @@
> #include <linux/ptrace.h>
> #include <asm/mmu.h>
> #include <asm/mce.h>
> +#include <asm/machdep.h>
>
> /* flush SLBs and reload */
> static void flush_and_reload_slb(void)
> @@ -197,13 +198,32 @@ static void mce_get_derror_p7(struct mce_error_info *mce_err, uint64_t dsisr)
> }
> }
>
> +static long mce_handle_ue_error(struct pt_regs *regs)
> +{
> + long handled = 0;
> +
> + /*
> + * On specific SCOM read via MMIO we may get a machine check
> + * exception with SRR0 pointing inside opal. If that is the
> + * case OPAL may have recovery address to re-read SCOM data in
> + * different way and hence we can recover from this MC.
> + */
> +
> + if (ppc_md.mce_check_early_recovery) {
> + if (ppc_md.mce_check_early_recovery(regs))
> + handled = 1;
> + }
> + return handled;
> +}
> +
> long __machine_check_early_realmode_p7(struct pt_regs *regs)
> {
> - uint64_t srr1, addr;
> + uint64_t srr1, nip, addr;
> long handled = 1;
> struct mce_error_info mce_error_info = { 0 };
>
> srr1 = regs->msr;
> + nip = regs->nip;
>
> /*
> * Handle memory errors depending whether this was a load/store or
> @@ -221,7 +241,11 @@ long __machine_check_early_realmode_p7(struct pt_regs *regs)
> addr = regs->nip;
> }
>
> - save_mce_event(regs, handled, &mce_error_info, addr);
> + /* Handle UE error. */
> + if (mce_error_info.error_type == MCE_ERROR_TYPE_UE)
> + handled = mce_handle_ue_error(regs);
> +
> + save_mce_event(regs, handled, &mce_error_info, nip, addr);
> return handled;
> }
>
> @@ -263,11 +287,12 @@ static long mce_handle_derror_p8(uint64_t dsisr)
>
> long __machine_check_early_realmode_p8(struct pt_regs *regs)
> {
> - uint64_t srr1, addr;
> + uint64_t srr1, nip, addr;
> long handled = 1;
> struct mce_error_info mce_error_info = { 0 };
>
> srr1 = regs->msr;
> + nip = regs->nip;
>
> if (P7_SRR1_MC_LOADSTORE(srr1)) {
> handled = mce_handle_derror_p8(regs->dsisr);
> @@ -279,6 +304,10 @@ long __machine_check_early_realmode_p8(struct pt_regs *regs)
> addr = regs->nip;
> }
>
> - save_mce_event(regs, handled, &mce_error_info, addr);
> + /* Handle UE error. */
> + if (mce_error_info.error_type == MCE_ERROR_TYPE_UE)
> + handled = mce_handle_ue_error(regs);
> +
> + save_mce_event(regs, handled, &mce_error_info, nip, addr);
> return handled;
> }
> diff --git a/arch/powerpc/kernel/prom.c b/arch/powerpc/kernel/prom.c
> index f58c0d3..d711b7e 100644
> --- a/arch/powerpc/kernel/prom.c
> +++ b/arch/powerpc/kernel/prom.c
> @@ -752,6 +752,11 @@ void __init early_init_devtree(void *params)
> spinning_secondaries = boot_cpu_count - 1;
> #endif
>
> +#ifdef CONFIG_PPC_POWERNV
> + /* Scan and build the list of machine check recoverable ranges */
> + of_scan_flat_dt(early_init_dt_scan_recoverable_ranges, NULL);
> +#endif
> +
> DBG(" <- early_init_devtree()\n");
> }
>
> diff --git a/arch/powerpc/platforms/powernv/opal.c b/arch/powerpc/platforms/powernv/opal.c
> index 65499ad..f52762b 100644
> --- a/arch/powerpc/platforms/powernv/opal.c
> +++ b/arch/powerpc/platforms/powernv/opal.c
> @@ -21,6 +21,7 @@
> #include <linux/sched.h>
> #include <linux/kobject.h>
> #include <linux/delay.h>
> +#include <linux/memblock.h>
> #include <asm/opal.h>
> #include <asm/firmware.h>
> #include <asm/mce.h>
> @@ -33,8 +34,18 @@ struct kobject *opal_kobj;
> struct opal {
> u64 base;
> u64 entry;
> + u64 size;
> } opal;
>
> +struct mcheck_recoverable_range {
> + u64 start_addr;
> + u64 end_addr;
> + u64 recover_addr;
> +};
> +
> +static struct mcheck_recoverable_range *mc_recoverable_range;
> +static int mc_recoverable_range_len;
> +
> static struct device_node *opal_node;
> static DEFINE_SPINLOCK(opal_write_lock);
> extern u64 opal_mc_secondary_handler[];
> @@ -49,25 +60,29 @@ static atomic_t opal_notifier_hold = ATOMIC_INIT(0);
> int __init early_init_dt_scan_opal(unsigned long node,
> const char *uname, int depth, void *data)
> {
> - const void *basep, *entryp;
> - unsigned long basesz, entrysz;
> + const void *basep, *entryp, *sizep;
> + unsigned long basesz, entrysz, runtimesz;
>
> if (depth != 1 || strcmp(uname, "ibm,opal") != 0)
> return 0;
>
> basep = of_get_flat_dt_prop(node, "opal-base-address", &basesz);
> entryp = of_get_flat_dt_prop(node, "opal-entry-address", &entrysz);
> + sizep = of_get_flat_dt_prop(node, "opal-runtime-size", &runtimesz);
>
> - if (!basep || !entryp)
> + if (!basep || !entryp || !sizep)
> return 1;
>
> opal.base = of_read_number(basep, basesz/4);
> opal.entry = of_read_number(entryp, entrysz/4);
> + opal.size = of_read_number(sizep, runtimesz/4);
>
> pr_debug("OPAL Base = 0x%llx (basep=%p basesz=%ld)\n",
> opal.base, basep, basesz);
> pr_debug("OPAL Entry = 0x%llx (entryp=%p basesz=%ld)\n",
> opal.entry, entryp, entrysz);
> + pr_debug("OPAL Entry = 0x%llx (sizep=%p runtimesz=%ld)\n",
> + opal.size, sizep, runtimesz);
>
> powerpc_firmware_features |= FW_FEATURE_OPAL;
> if (of_flat_dt_is_compatible(node, "ibm,opal-v3")) {
> @@ -84,6 +99,65 @@ int __init early_init_dt_scan_opal(unsigned long node,
> return 1;
> }
>
> +int __init early_init_dt_scan_recoverable_ranges(unsigned long node,
> + const char *uname, int depth, void *data)
> +{
> + unsigned long i, psize, size;
> + const __be32 *prop;
> +
> + if (depth != 1 || strcmp(uname, "ibm,opal") != 0)
> + return 0;
> +
> + prop = of_get_flat_dt_prop(node, "mcheck-recoverable-ranges", &psize);
> +
> + if (!prop)
> + return 1;
> +
> + pr_debug("Found machine check recoverable ranges.\n");
> +
> + /*
> + * Calculate number of available entries.
> + *
> + * Each recoverable address range entry is (start address, len,
> + * recovery address), 2 cells each for start and recovery address,
> + * 1 cell for len, totalling 5 cells per entry.
> + */
> + mc_recoverable_range_len = psize / (sizeof(*prop) * 5);
> +
> + /* Sanity check */
> + if (!mc_recoverable_range_len)
> + return 1;
> +
> + /* Size required to hold all the entries. */
> + size = mc_recoverable_range_len *
> + sizeof(struct mcheck_recoverable_range);
> +
> + /*
> + * Allocate a buffer to hold the MC recoverable ranges. We would be
> + * accessing them in real mode, hence it needs to be within
> + * RMO region.
> + */
> + mc_recoverable_range = __va(memblock_alloc_base(size, __alignof__(u64),
> + ppc64_rma_size));
> + memset(mc_recoverable_range, 0, size);
> +
> + for (i = 0; i < mc_recoverable_range_len; i++) {
> + mc_recoverable_range[i].start_addr =
> + of_read_number(prop + (i * 5) + 0, 2);
> + mc_recoverable_range[i].end_addr =
> + mc_recoverable_range[i].start_addr +
> + of_read_number(prop + (i * 5) + 2, 1);
> + mc_recoverable_range[i].recover_addr =
> + of_read_number(prop + (i * 5) + 3, 2);
> +
> + pr_debug("Machine check recoverable range: %llx..%llx: %llx\n",
> + mc_recoverable_range[i].start_addr,
> + mc_recoverable_range[i].end_addr,
> + mc_recoverable_range[i].recover_addr);
> + }
> + return 1;
> +}
> +
> static int __init opal_register_exception_handlers(void)
> {
> #ifdef __BIG_ENDIAN__
> @@ -401,6 +475,38 @@ int opal_machine_check(struct pt_regs *regs)
> return 0;
> }
>
> +static uint64_t find_recovery_address(uint64_t nip)
> +{
> + int i;
> +
> + for (i = 0; i < mc_recoverable_range_len; i++)
> + if ((nip >= mc_recoverable_range[i].start_addr) &&
> + (nip < mc_recoverable_range[i].end_addr))
> + return mc_recoverable_range[i].recover_addr;
> + return 0;
> +}
> +
> +bool opal_mce_check_early_recovery(struct pt_regs *regs)
> +{
> + uint64_t recover_addr = 0;
> +
> + if (!opal.base || !opal.size)
> + goto out;
> +
> + if ((regs->nip >= opal.base) &&
> + (regs->nip <= (opal.base + opal.size)))
> + recover_addr = find_recovery_address(regs->nip);
> +
> + /*
> + * Setup regs->nip to rfi into fixup address.
> + */
> + if (recover_addr)
> + regs->nip = recover_addr;
> +
> +out:
> + return !!recover_addr;
> +}
> +
> static irqreturn_t opal_interrupt(int irq, void *data)
> {
> __be64 events;
> diff --git a/arch/powerpc/platforms/powernv/setup.c b/arch/powerpc/platforms/powernv/setup.c
> index 110f4fb..2d80845 100644
> --- a/arch/powerpc/platforms/powernv/setup.c
> +++ b/arch/powerpc/platforms/powernv/setup.c
> @@ -188,6 +188,7 @@ static void __init pnv_setup_machdep_opal(void)
> ppc_md.power_off = pnv_power_off;
> ppc_md.halt = pnv_halt;
> ppc_md.machine_check_exception = opal_machine_check;
> + ppc_md.mce_check_early_recovery = opal_mce_check_early_recovery;
> }
>
> #ifdef CONFIG_PPC_POWERNV_RTAS
More information about the Linuxppc-dev
mailing list