From: Hari Bathini <hbath...@linux.vnet.ibm.com>
Firmware provides architected register state data at the time of crash.
Process this data and build CPU notes to append to ELF core.
Signed-off-by: Hari Bathini <hbath...@linux.vnet.ibm.com>
Signed-off-by: Vasant Hegde <hegdevas...@linux.vnet.ibm.com>
---
Changes in v2:
* Updated reg type values according to recent OPAL changes
arch/powerpc/include/asm/opal-api.h | 23 +++
arch/powerpc/kernel/fadump-common.h | 3
arch/powerpc/platforms/powernv/opal-fadump.c | 187 ++++++++++++++++++++++++--
arch/powerpc/platforms/powernv/opal-fadump.h | 4 +
4 files changed, 206 insertions(+), 11 deletions(-)
diff --git a/arch/powerpc/include/asm/opal-api.h
b/arch/powerpc/include/asm/opal-api.h
index 75471c2..91f2735 100644
--- a/arch/powerpc/include/asm/opal-api.h
+++ b/arch/powerpc/include/asm/opal-api.h
@@ -976,6 +976,29 @@ struct opal_sg_list {
* Firmware-Assisted Dump (FADump)
*/
+/* FADump thread header for register entries */
+struct opal_fadump_thread_hdr {
+ __be32 pir;
+ /* 0x00 - 0x0F - The corresponding stop state of the core */
+ u8 core_state;
+ u8 reserved[3];
+
+ __be32 offset; /* Offset to Register Entries array */
+ __be32 ecnt; /* Number of entries */
+ __be32 esize; /* Alloc size of each array entry in bytes */
+ __be32 eactsz; /* Actual size of each array entry in bytes */
+} __packed;
+
+#define OPAL_REG_TYPE_GPR 0x01
+#define OPAL_REG_TYPE_SPR 0x02
+
+/* FADump register entry. */
+struct opal_fadump_reg_entry {
+ __be32 reg_type;
+ __be32 reg_num;
+ __be64 reg_val;
+};
+
/* The maximum number of dump sections supported by OPAL */
#define OPAL_FADUMP_NR_SECTIONS 64
diff --git a/arch/powerpc/kernel/fadump-common.h
b/arch/powerpc/kernel/fadump-common.h
index ff764d4..8d47382 100644
--- a/arch/powerpc/kernel/fadump-common.h
+++ b/arch/powerpc/kernel/fadump-common.h
@@ -117,6 +117,9 @@ struct fadump_memory_range {
/* Firmware-assisted dump configuration details. */
struct fw_dump {
+ unsigned long cpu_state_destination_addr;
+ unsigned long cpu_state_data_version;
+ unsigned long cpu_state_entry_size;
unsigned long cpu_state_data_size;
unsigned long hpte_region_size;
unsigned long boot_memory_size;
diff --git a/arch/powerpc/platforms/powernv/opal-fadump.c
b/arch/powerpc/platforms/powernv/opal-fadump.c
index da8480d..853f663 100644
--- a/arch/powerpc/platforms/powernv/opal-fadump.c
+++ b/arch/powerpc/platforms/powernv/opal-fadump.c
@@ -94,6 +94,12 @@ static void update_fadump_config(struct fw_dump *fadump_conf,
last_end = base + size;
j++;
+ } else if (fdm->section[i].src_type ==
+ OPAL_FADUMP_CPU_STATE_DATA) {
+ fadump_conf->cpu_state_destination_addr =
+ be64_to_cpu(fdm->section[i].dest_addr);
+ fadump_conf->cpu_state_data_size =
+ be64_to_cpu(fdm->section[i].dest_size);
}
}
fadump_conf->rmr_regions_cnt = j;
@@ -199,6 +205,75 @@ static int opal_invalidate_fadump(struct fw_dump
*fadump_conf)
return 0;
}
+static inline void fadump_set_regval_regnum(struct pt_regs *regs, u32 reg_type,
+ u32 reg_num, u64 reg_val)
+{
+ if (reg_type == OPAL_REG_TYPE_GPR) {
+ if (reg_num < 32)
+ regs->gpr[reg_num] = reg_val;
+ return;
+ }
+
+ switch (reg_num) {
+ case 2000:
+ regs->nip = reg_val;
+ break;
+ case 2001:
+ regs->msr = reg_val;
+ break;
+ case 9:
+ regs->ctr = reg_val;
+ break;
+ case 8:
+ regs->link = reg_val;
+ break;
+ case 1:
+ regs->xer = reg_val;
+ break;
+ case 2002:
+ regs->ccr = reg_val;
+ break;
+ case 19:
+ regs->dar = reg_val;
+ break;
+ case 18:
+ regs->dsisr = reg_val;
+ break;
+ }
+}
+
+static inline void fadump_read_registers(char *bufp, unsigned int regs_cnt,
+ unsigned int reg_entry_size,
+ struct pt_regs *regs)
+{
+ int i;
+ struct opal_fadump_reg_entry *reg_entry;
+
+ memset(regs, 0, sizeof(struct pt_regs));
+
+ for (i = 0; i < regs_cnt; i++, bufp += reg_entry_size) {
+ reg_entry = (struct opal_fadump_reg_entry *)bufp;
+ fadump_set_regval_regnum(regs,
+ be32_to_cpu(reg_entry->reg_type),
+ be32_to_cpu(reg_entry->reg_num),
+ be64_to_cpu(reg_entry->reg_val));
+ }
+}
+
+static inline bool __init is_thread_core_inactive(u8 core_state)
+{
+ bool is_inactive = false;
+
+ /*
+ * TODO: Check what each state value means and implement accordingly.
+ * Also, make sure F/W is updating this field appropriately.
+ */
+ if (core_state == CORE_STATE_INACTIVE)
+ is_inactive = true;
+
+ return is_inactive;
+}
+
/*
* Read CPU state dump data and convert it into ELF notes.
*
@@ -210,8 +285,32 @@ static int __init fadump_build_cpu_notes(struct fw_dump
*fadump_conf)
{
u32 num_cpus, *note_buf;
struct fadump_crash_info_header *fdh = NULL;
+ struct opal_fadump_thread_hdr *thdr;
+ unsigned long addr;
+ u32 thread_pir;
+ char *bufp;
+ struct pt_regs regs;
+ unsigned int size_of_each_thread;
+ unsigned int regs_offset, regs_cnt, reg_esize;
+ int i;
+
+ size_of_each_thread = fadump_conf->cpu_state_entry_size;
+ num_cpus = (fadump_conf->cpu_state_data_size / size_of_each_thread);
+
+ addr = fadump_conf->cpu_state_destination_addr;
+ bufp = __va(addr);
+
+ /*
+ * Offset for register entries, entry size and registers count is
+ * duplicated in every thread header in keeping with HDAT format.
+ * Use these values from the first thread header.
+ */
+ thdr = (struct opal_fadump_thread_hdr *)bufp;
+ regs_offset = (offsetof(struct opal_fadump_thread_hdr, offset) +
+ be32_to_cpu(thdr->offset));
+ reg_esize = be32_to_cpu(thdr->esize);
+ regs_cnt = be32_to_cpu(thdr->ecnt);
- num_cpus = 1;
/* Allocate buffer to hold cpu crash notes. */
fadump_conf->cpu_notes_buf_size = num_cpus * sizeof(note_buf_t);
fadump_conf->cpu_notes_buf_size =
@@ -230,10 +329,53 @@ static int __init fadump_build_cpu_notes(struct fw_dump
*fadump_conf)
if (fadump_conf->fadumphdr_addr)
fdh = __va(fadump_conf->fadumphdr_addr);
- if (fdh && (fdh->crashing_cpu != CPU_UNKNOWN)) {
- note_buf = fadump_regs_to_elf_notes(note_buf, &(fdh->regs));
- final_note(note_buf);
+ pr_debug("--------CPU State Data------------\n");
+ pr_debug("NumCpus : %u\n", num_cpus);
+ pr_debug("\tOffset: %u, Entry size: %u, Cnt: %u\n",
+ regs_offset, reg_esize, regs_cnt);
+
+ for (i = 0; i < num_cpus; i++, bufp += size_of_each_thread) {
+ thdr = (struct opal_fadump_thread_hdr *)bufp;
+
+ thread_pir = be32_to_cpu(thdr->pir);
+ pr_debug("%04d) PIR: 0x%x, core state: 0x%02x\n",
+ (i + 1), thread_pir, thdr->core_state);
+
+ /*
+ * Register state data of MAX cores is provided by firmware,
+ * but some of this cores may not be active. So, while
+ * processing register state data, check core state and
+ * skip threads that belong to inactive cores.
+ */
+ if (is_thread_core_inactive(thdr->core_state))
+ continue;
+
+ /*
+ * If this is kernel initiated crash, crashing_cpu would be set
+ * appropriately and register data of the crashing CPU saved by
+ * crashing kernel. Add this saved register data of crashing CPU
+ * to elf notes and populate the pt_regs for the remaining CPUs
+ * from register state data provided by firmware.
+ */
+ if (fdh && (fdh->crashing_cpu == thread_pir)) {
+ note_buf = fadump_regs_to_elf_notes(note_buf,
+ &fdh->regs);
+ pr_debug("Crashing CPU PIR: 0x%x - R1 : 0x%lx, NIP :
0x%lx\n",
+ fdh->crashing_cpu, fdh->regs.gpr[1],
+ fdh->regs.nip);
+ continue;
+ }
+
+ fadump_read_registers((bufp + regs_offset), regs_cnt,
+ reg_esize, ®s);
+
+ note_buf = fadump_regs_to_elf_notes(note_buf, ®s);
+ pr_debug("CPU PIR: 0x%x - R1 : 0x%lx, NIP : 0x%lx\n",
+ thread_pir, regs.gpr[1], regs.nip);
+ }
+ final_note(note_buf);
+ if (fdh) {
pr_debug("Updating elfcore header (%llx) with cpu notes\n",
fdh->elfcorehdr_addr);
fadump_update_elfcore_header(fadump_conf,
@@ -258,13 +400,6 @@ static int __init opal_process_fadump(struct fw_dump
*fadump_conf)
return -EINVAL;
}
- /*
- * TODO: To build cpu notes, find a way to map PIR to logical id.
- * Also, we may need different method for pseries and powernv.
- * The currently booted kernel could have a different PIR to
- * logical id mapping. So, try saving info of previous kernel's
- * paca to get the right PIR to logical id mapping.
- */
rc = fadump_build_cpu_notes(fadump_conf);
if (rc)
return rc;
@@ -319,6 +454,14 @@ static void opal_crash_fadump(struct
fadump_crash_info_header *fdh,
{
int rc;
+ /*
+ * Unlike on pseries platform, logical CPU number is not provided
+ * with architected register state data. So, store the crashing
+ * CPU's PIR instead to plug the appropriate register data for
+ * crashing CPU in the vmcore file.
+ */
+ fdh->crashing_cpu = (u32)mfspr(SPRN_PIR);
+
rc = opal_cec_reboot2(OPAL_REBOOT_OS_ERROR, msg);
if (rc == OPAL_UNSUPPORTED) {
pr_emerg("Reboot type %d not supported.\n",
@@ -340,6 +483,7 @@ static struct fadump_ops opal_fadump_ops = {
int __init opal_dt_scan_fadump(struct fw_dump *fadump_conf, ulong node)
{
unsigned long dn;
+ const __be32 *prop;
/*
* Check if Firmware-Assisted Dump is supported. if yes, check
@@ -351,6 +495,17 @@ int __init opal_dt_scan_fadump(struct fw_dump
*fadump_conf, ulong node)
return 1;
}
+ prop = of_get_flat_dt_prop(dn, "cpu-data-version", NULL);
+ if (prop)
+ fadump_conf->cpu_state_data_version = of_read_number(prop, 1);
+
+ if (fadump_conf->cpu_state_data_version != CPU_STATE_DATA_VERSION) {
+ pr_err("CPU state data format version mismatch!\n");
+ pr_err("Kernel: %u, OPAL: %lu\n", CPU_STATE_DATA_VERSION,
+ fadump_conf->cpu_state_data_version);
+ return 1;
+ }
+
/*
* Firmware currently supports only 32-bit value for size,
* align it to 1MB size.
@@ -365,6 +520,16 @@ int __init opal_dt_scan_fadump(struct fw_dump
*fadump_conf, ulong node)
pr_info("Firmware-assisted dump is active.\n");
fadump_conf->dump_active = 1;
update_fadump_config(fadump_conf, (void *)__pa(fdm_active));
+
+ /*
+ * Doesn't need to populate these fields while registering dump
+ * as destination address and size are provided by F/W.
+ */
+ prop = of_get_flat_dt_prop(dn, "cpu-data-size", NULL);
+ if (prop) {
+ fadump_conf->cpu_state_entry_size =
+ of_read_number(prop, 1);
+ }
}
fadump_conf->ops = &opal_fadump_ops;
diff --git a/arch/powerpc/platforms/powernv/opal-fadump.h
b/arch/powerpc/platforms/powernv/opal-fadump.h
index a5eeb2c..80aee2d 100644
--- a/arch/powerpc/platforms/powernv/opal-fadump.h
+++ b/arch/powerpc/platforms/powernv/opal-fadump.h
@@ -13,6 +13,8 @@
#ifndef __PPC64_OPAL_FA_DUMP_H__
#define __PPC64_OPAL_FA_DUMP_H__
+#define CPU_STATE_DATA_VERSION 1
+
#define OPAL_FADUMP_CPU_STATE_DATA 0x0000
/* OPAL : 0x01 – 0x39 */
#define OPAL_FADUMP_OPAL_REGION 0x0001
@@ -37,4 +39,6 @@ enum opal_fadump_section_types {
#define OPAL_MAX_SECTIONS (OPAL_SECTIONS + \
MAX_REAL_MEM_REGIONS - 1)
+#define CORE_STATE_INACTIVE (0x0F)
+
#endif /* __PPC64_OPAL_FA_DUMP_H__ */
