[PATCH v3 04/11] pstore/blk: pstore/zone: support pmsg recorder

From: WeiXiong Liao
Date: Wed Mar 25 2020 - 04:56:06 EST

pmsg support recorder for userspace. To enable pmsg, just make pmsg_size
be greater than 0 and a multiple of 4096.

Signed-off-by: WeiXiong Liao <liaoweixiong@xxxxxxxxxxxxxxxxx>
---
fs/pstore/Kconfig | 12 ++
fs/pstore/pstore_blk.c | 9 ++
fs/pstore/pstore_zone.c | 273 +++++++++++++++++++++++++++++++++++++++++---
include/linux/pstore_zone.h | 2 +
4 files changed, 283 insertions(+), 13 deletions(-)

diff --git a/fs/pstore/Kconfig b/fs/pstore/Kconfig
index 590af61019c2..8cead860dcfc 100644
--- a/fs/pstore/Kconfig
+++ b/fs/pstore/Kconfig
@@ -213,6 +213,18 @@ config PSTORE_BLK_OOPS_SIZE
NOTE that, both Kconfig and module parameters can configure
pstore/blk, but module parameters have priority over Kconfig.

+config PSTORE_BLK_PMSG_SIZE
+ int "Size in Kbytes of pmsg to store"
+ depends on PSTORE_BLK
+ depends on PSTORE_PMSG
+ default 64
+ help
+ This just sets size of pmsg (pmsg_size) for pstore/blk. The size is
+ in KB and must be a multiple of 4.
+
+ NOTE that, both Kconfig and module parameters can configure
+ pstore/blk, but module parameters have priority over Kconfig.
+
config PSTORE_BLK_DUMP_OOPS
bool "dump oops"
depends on PSTORE_BLK
diff --git a/fs/pstore/pstore_blk.c b/fs/pstore/pstore_blk.c
index f3ce7bbd9077..85cd9f2335be 100644
--- a/fs/pstore/pstore_blk.c
+++ b/fs/pstore/pstore_blk.c
@@ -18,6 +18,14 @@
module_param(oops_size, long, 0400);
MODULE_PARM_DESC(oops_size, "oops size in kbytes");

+#if IS_ENABLED(CONFIG_PSTORE_PMSG)
+static long pmsg_size = CONFIG_PSTORE_BLK_PMSG_SIZE;
+#else
+static long pmsg_size = -1;
+#endif
+module_param(pmsg_size, long, 0400);
+MODULE_PARM_DESC(pmsg_size, "pmsg size in kbytes");
+
static int dump_oops = CONFIG_PSTORE_BLK_DUMP_OOPS;
module_param(dump_oops, int, 0400);
MODULE_PARM_DESC(total_size, "whether dump oops");
@@ -122,6 +130,7 @@ static int psblk_register_do(struct psblk_device *dev)
}

verify_size(oops_size, 4096, dev->flags & PSTORE_FLAGS_DMESG);
+ verify_size(pmsg_size, 4096, dev->flags & PSTORE_FLAGS_PMSG);
#undef verify_size
dump_oops = dump_oops <= 0 ? 0 : 1;

diff --git a/fs/pstore/pstore_zone.c b/fs/pstore/pstore_zone.c
index 62c834072498..444bce7f9ac3 100644
--- a/fs/pstore/pstore_zone.c
+++ b/fs/pstore/pstore_zone.c
@@ -23,12 +23,14 @@
*
* @sig: signature to indicate header (PSZ_SIG xor PSZONE-type value)
* @datalen: length of data in @data
+ * @start: offset into @data where the beginning of the stored bytes begin
* @data: zone data.
*/
struct psz_buffer {
#define PSZ_SIG (0x43474244) /* DBGC */
uint32_t sig;
atomic_t datalen;
+ atomic_t start;
uint8_t data[];
};

@@ -84,9 +86,11 @@ struct psz_zone {
* struct psz_context - all about running state of pstore/zone
*
* @opszs: oops/panic storage zones
+ * @ppsz: pmsg storage zone
* @oops_max_cnt: max count of @opszs
* @oops_read_cnt: counter to read oops zone
* @oops_write_cnt: counter to write
+ * @pmsg_read_cnt: counter to read pmsg zone
* @oops_counter: counter to oops
* @panic_counter: counter to panic
* @recovered: whether finish recovering data from storage
@@ -97,9 +101,11 @@ struct psz_zone {
*/
struct psz_context {
struct psz_zone **opszs;
+ struct psz_zone *ppsz;
unsigned int oops_max_cnt;
unsigned int oops_read_cnt;
unsigned int oops_write_cnt;
+ unsigned int pmsg_read_cnt;
/*
* the counter should be recovered when recover.
* It records the oops/panic times after burning rather than booting.
@@ -139,6 +145,11 @@ static inline int buffer_datalen(struct psz_zone *zone)
return atomic_read(&zone->buffer->datalen);
}

+static inline int buffer_start(struct psz_zone *zone)
+{
+ return atomic_read(&zone->buffer->start);
+}
+
static inline bool is_on_panic(void)
{
struct psz_context *cxt = &psz_cxt;
@@ -146,10 +157,10 @@ static inline bool is_on_panic(void)
return atomic_read(&cxt->on_panic);
}

-static ssize_t psz_zone_read(struct psz_zone *zone, char *buf,
+static ssize_t psz_zone_read_buffer(struct psz_zone *zone, char *buf,
size_t len, unsigned long off)
{
- if (!buf || !zone->buffer)
+ if (!buf || !zone || !zone->buffer)
return -EINVAL;
if (off > zone->buffer_size)
return -EINVAL;
@@ -158,6 +169,18 @@ static ssize_t psz_zone_read(struct psz_zone *zone, char *buf,
return len;
}

+static int psz_zone_read_oldbuf(struct psz_zone *zone, char *buf,
+ size_t len, unsigned long off)
+{
+ if (!buf || !zone || !zone->oldbuf)
+ return -EINVAL;
+ if (off > zone->buffer_size)
+ return -EINVAL;
+ len = min_t(size_t, len, zone->buffer_size - off);
+ memcpy(buf, zone->oldbuf->data + off, len);
+ return 0;
+}
+
static int psz_zone_write(struct psz_zone *zone,
enum psz_flush_mode flush_mode, const char *buf,
size_t len, unsigned long off)
@@ -413,6 +436,93 @@ static int psz_recover_oops(struct psz_context *cxt)
return ret;
}

+static int psz_recover_zone(struct psz_context *cxt, struct psz_zone *zone)
+{
+ struct psz_info *info = cxt->psz_info;
+ struct psz_buffer *oldbuf, tmpbuf;
+ int ret = 0;
+ char *buf;
+ ssize_t rcnt, len, start, off;
+
+ if (!zone || zone->oldbuf)
+ return 0;
+
+ if (is_on_panic()) {
+ /* save data as much as possible */
+ psz_flush_dirty_zone(zone);
+ return 0;
+ }
+
+ if (unlikely(!info->read))
+ return -EINVAL;
+
+ len = sizeof(struct psz_buffer);
+ rcnt = info->read((char *)&tmpbuf, len, zone->off);
+ if (rcnt != len) {
+ pr_debug("read zone %s failed\n", zone->name);
+ return (int)rcnt < 0 ? (int)rcnt : -EIO;
+ }
+
+ if (tmpbuf.sig != zone->buffer->sig) {
+ pr_debug("no valid data in zone %s\n", zone->name);
+ return 0;
+ }
+
+ if (zone->buffer_size < atomic_read(&tmpbuf.datalen) ||
+ zone->buffer_size < atomic_read(&tmpbuf.start)) {
+ pr_info("found overtop zone: %s: off %lld, size %zu\n",
+ zone->name, zone->off, zone->buffer_size);
+ /* just keep going */
+ return 0;
+ }
+
+ if (!atomic_read(&tmpbuf.datalen)) {
+ pr_debug("found erased zone: %s: off %lld, size %zu, datalen %d\n",
+ zone->name, zone->off, zone->buffer_size,
+ atomic_read(&tmpbuf.datalen));
+ return 0;
+ }
+
+ pr_debug("found nice zone: %s: off %lld, size %zu, datalen %d\n",
+ zone->name, zone->off, zone->buffer_size,
+ atomic_read(&tmpbuf.datalen));
+
+ len = atomic_read(&tmpbuf.datalen) + sizeof(*oldbuf);
+ oldbuf = kzalloc(len, GFP_KERNEL);
+ if (!oldbuf)
+ return -ENOMEM;
+
+ memcpy(oldbuf, &tmpbuf, sizeof(*oldbuf));
+ buf = (char *)oldbuf + sizeof(*oldbuf);
+ len = atomic_read(&oldbuf->datalen);
+ start = atomic_read(&oldbuf->start);
+ off = zone->off + sizeof(*oldbuf);
+
+ /* get part of data */
+ rcnt = info->read(buf, len - start, off + start);
+ if (rcnt != len - start) {
+ pr_err("read zone %s failed\n", zone->name);
+ ret = (int)rcnt < 0 ? (int)rcnt : -EIO;
+ goto free_oldbuf;
+ }
+
+ /* get the rest of data */
+ rcnt = info->read(buf + len - start, start, off);
+ if (rcnt != start) {
+ pr_err("read zone %s failed\n", zone->name);
+ ret = (int)rcnt < 0 ? (int)rcnt : -EIO;
+ goto free_oldbuf;
+ }
+
+ zone->oldbuf = oldbuf;
+ psz_flush_dirty_zone(zone);
+ return 0;
+
+free_oldbuf:
+ kfree(oldbuf);
+ return ret;
+}
+
/**
* psz_recovery() - recover data from storage
* @cxt: the context of pstore/zone
@@ -432,6 +542,10 @@ static inline int psz_recovery(struct psz_context *cxt)
if (ret)
goto recover_fail;

+ ret = psz_recover_zone(cxt, cxt->ppsz);
+ if (ret)
+ goto recover_fail;
+
pr_debug("recover end!\n");
atomic_set(&cxt->recovered, 1);
return 0;
@@ -446,9 +560,17 @@ static int psz_pstore_open(struct pstore_info *psi)
struct psz_context *cxt = psi->data;

cxt->oops_read_cnt = 0;
+ cxt->pmsg_read_cnt = 0;
return 0;
}

+static inline bool psz_old_ok(struct psz_zone *zone)
+{
+ if (zone && zone->oldbuf && atomic_read(&zone->oldbuf->datalen))
+ return true;
+ return false;
+}
+
static inline bool psz_ok(struct psz_zone *zone)
{
if (zone && zone->buffer && buffer_datalen(zone))
@@ -473,6 +595,25 @@ static inline int psz_oops_erase(struct psz_context *cxt,
return psz_zone_write(zone, FLUSH_META, NULL, 0, 0);
}

+static inline int psz_record_erase(struct psz_context *cxt,
+ struct psz_zone *zone)
+{
+ if (unlikely(!psz_old_ok(zone)))
+ return 0;
+
+ kfree(zone->oldbuf);
+ zone->oldbuf = NULL;
+ /*
+ * if there are new data in zone buffer, that means the old data
+ * are already invalid. It is no need to flush 0 (erase) to
+ * block device.
+ */
+ if (!buffer_datalen(zone))
+ return psz_zone_write(zone, FLUSH_META, NULL, 0, 0);
+ psz_flush_dirty_zone(zone);
+ return 0;
+}
+
static int psz_pstore_erase(struct pstore_record *record)
{
struct psz_context *cxt = record->psi->data;
@@ -482,6 +623,8 @@ static int psz_pstore_erase(struct pstore_record *record)
if (record->id >= cxt->oops_max_cnt)
return -EINVAL;
return psz_oops_erase(cxt, cxt->opszs[record->id], record);
+ case PSTORE_TYPE_PMSG:
+ return psz_record_erase(cxt, cxt->ppsz);
default:
return -EINVAL;
}
@@ -502,8 +645,10 @@ static void psz_write_kmsg_hdr(struct psz_zone *zone,
hdr->reason = record->reason;
if (hdr->reason == KMSG_DUMP_OOPS)
hdr->counter = ++cxt->oops_counter;
- else
+ else if (hdr->reason == KMSG_DUMP_PANIC)
hdr->counter = ++cxt->panic_counter;
+ else
+ hdr->counter = 0;
}

static inline int notrace psz_oops_write_record(struct psz_context *cxt,
@@ -567,6 +712,53 @@ static int notrace psz_oops_write(struct psz_context *cxt,
return 0;
}

+static int notrace psz_record_write(struct psz_zone *zone,
+ struct pstore_record *record)
+{
+ size_t start, rem;
+ int cnt = record->size;
+ bool is_full_data = false;
+ char *buf = record->buf;
+
+ if (!zone || !record)
+ return -ENOSPC;
+
+ if (atomic_read(&zone->buffer->datalen) >= zone->buffer_size)
+ is_full_data = true;
+
+ if (unlikely(cnt > zone->buffer_size)) {
+ buf += cnt - zone->buffer_size;
+ cnt = zone->buffer_size;
+ }
+
+ start = buffer_start(zone);
+ rem = zone->buffer_size - start;
+ if (unlikely(rem < cnt)) {
+ psz_zone_write(zone, FLUSH_PART, buf, rem, start);
+ buf += rem;
+ cnt -= rem;
+ start = 0;
+ is_full_data = true;
+ }
+
+ atomic_set(&zone->buffer->start, cnt + start);
+ psz_zone_write(zone, FLUSH_PART, buf, cnt, start);
+
+ /**
+ * psz_zone_write will set datalen as start + cnt.
+ * It work if actual data length lesser than buffer size.
+ * If data length greater than buffer size, pmsg will rewrite to
+ * beginning of zone, which make buffer->datalen wrongly.
+ * So we should reset datalen as buffer size once actual data length
+ * greater than buffer size.
+ */
+ if (is_full_data) {
+ atomic_set(&zone->buffer->datalen, zone->buffer_size);
+ psz_zone_write(zone, FLUSH_META, NULL, 0, 0);
+ }
+ return 0;
+}
+
static int notrace psz_pstore_write(struct pstore_record *record)
{
struct psz_context *cxt = record->psi->data;
@@ -578,6 +770,8 @@ static int notrace psz_pstore_write(struct pstore_record *record)
switch (record->type) {
case PSTORE_TYPE_DMESG:
return psz_oops_write(cxt, record);
+ case PSTORE_TYPE_PMSG:
+ return psz_record_write(cxt->ppsz, record);
default:
return -EINVAL;
}
@@ -593,6 +787,13 @@ static struct psz_zone *psz_read_next_zone(struct psz_context *cxt)
return zone;
}

+ if (cxt->pmsg_read_cnt == 0) {
+ cxt->pmsg_read_cnt++;
+ zone = cxt->ppsz;
+ if (psz_old_ok(zone))
+ return zone;
+ }
+
return NULL;
}

@@ -642,7 +843,7 @@ static ssize_t psz_oops_read(struct psz_zone *zone,
return -ENOMEM;
}

- size = psz_zone_read(zone, record->buf + hlen, size,
+ size = psz_zone_read_buffer(zone, record->buf + hlen, size,
sizeof(struct psz_oops_header) < 0);
if (unlikely(size < 0)) {
kfree(record->buf);
@@ -652,6 +853,32 @@ static ssize_t psz_oops_read(struct psz_zone *zone,
return size + hlen;
}

+static ssize_t psz_record_read(struct psz_zone *zone,
+ struct pstore_record *record)
+{
+ size_t len;
+ struct psz_buffer *buf;
+
+ if (!zone || !record)
+ return -ENOSPC;
+
+ buf = (struct psz_buffer *)zone->oldbuf;
+ if (!buf)
+ return -ENOMSG;
+
+ len = atomic_read(&buf->datalen);
+ record->buf = kmalloc(len, GFP_KERNEL);
+ if (!record->buf)
+ return -ENOMEM;
+
+ if (unlikely(psz_zone_read_oldbuf(zone, record->buf, len, 0))) {
+ kfree(record->buf);
+ return -ENOMSG;
+ }
+
+ return len;
+}
+
static ssize_t psz_pstore_read(struct pstore_record *record)
{
struct psz_context *cxt = record->psi->data;
@@ -676,6 +903,9 @@ static ssize_t psz_pstore_read(struct pstore_record *record)
readop = psz_oops_read;
record->id = cxt->oops_read_cnt - 1;
break;
+ case PSTORE_TYPE_PMSG:
+ readop = psz_record_read;
+ break;
default:
goto next_zone;
}
@@ -731,8 +961,10 @@ static struct psz_zone *psz_init_zone(enum pstore_type_id type,
zone->type = type;
zone->buffer_size = size - sizeof(struct psz_buffer);
zone->buffer->sig = type ^ PSZ_SIG;
+ zone->oldbuf = NULL;
atomic_set(&zone->dirty, 0);
atomic_set(&zone->buffer->datalen, 0);
+ atomic_set(&zone->buffer->start, 0);

*off += size;

@@ -816,6 +1048,8 @@ static void psz_free_all_zones(struct psz_context *cxt)
{
if (cxt->opszs)
psz_free_zones(&cxt->opszs, &cxt->oops_max_cnt);
+ if (cxt->ppsz)
+ psz_free_zone(&cxt->ppsz);
}

static int psz_alloc_zones(struct psz_context *cxt)
@@ -823,18 +1057,26 @@ static int psz_alloc_zones(struct psz_context *cxt)
struct psz_info *info = cxt->psz_info;
loff_t off = 0;
int err;
- size_t size;
+ size_t off_size = 0;
+
+ off_size += info->pmsg_size;
+ cxt->ppsz = psz_init_zone(PSTORE_TYPE_PMSG, &off, info->pmsg_size);
+ if (IS_ERR(cxt->ppsz)) {
+ err = PTR_ERR(cxt->ppsz);
+ goto free_out;
+ }

- size = info->total_size;
- cxt->opszs = psz_init_zones(PSTORE_TYPE_DMESG, &off, size,
+ cxt->opszs = psz_init_zones(PSTORE_TYPE_DMESG, &off,
+ info->total_size - off_size,
info->oops_size, &cxt->oops_max_cnt);
if (IS_ERR(cxt->opszs)) {
err = PTR_ERR(cxt->opszs);
- goto fail_out;
+ goto free_out;
}

return 0;
-fail_out:
+free_out:
+ psz_free_all_zones(cxt);
return err;
}

@@ -858,7 +1100,7 @@ int psz_register(struct psz_info *info)
return -EINVAL;
}

- if (!info->oops_size) {
+ if (!info->oops_size && !info->pmsg_size) {
pr_warn("at least one of the records be non-zero\n");
return -EINVAL;
}
@@ -885,6 +1127,7 @@ int psz_register(struct psz_info *info)

check_size(total_size, 4096);
check_size(oops_size, SECTOR_SIZE);
+ check_size(pmsg_size, SECTOR_SIZE);

#undef check_size

@@ -916,6 +1159,7 @@ int psz_register(struct psz_info *info)
pr_debug("register %s with properties:\n", info->name);
pr_debug("\ttotal size : %ld Bytes\n", info->total_size);
pr_debug("\toops size : %ld Bytes\n", info->oops_size);
+ pr_debug("\tpmsg size : %ld Bytes\n", info->pmsg_size);

err = psz_alloc_zones(cxt);
if (err) {
@@ -934,11 +1178,14 @@ int psz_register(struct psz_info *info)
}
cxt->pstore.data = cxt;
if (info->oops_size)
- cxt->pstore.flags = PSTORE_FLAGS_DMESG;
+ cxt->pstore.flags |= PSTORE_FLAGS_DMESG;
+ if (info->pmsg_size)
+ cxt->pstore.flags |= PSTORE_FLAGS_PMSG;

- pr_info("Registered %s as pszone backend for%s%s\n", info->name,
+ pr_info("Registered %s as pszone backend for%s%s%s\n", info->name,
cxt->opszs && cxt->psz_info->dump_oops ? " Oops" : "",
- cxt->opszs && cxt->psz_info->panic_write ? " Panic" : "");
+ cxt->opszs && cxt->psz_info->panic_write ? " Panic" : "",
+ cxt->ppsz ? " Pmsg" : "");

err = pstore_register(&cxt->pstore);
if (err) {
diff --git a/include/linux/pstore_zone.h b/include/linux/pstore_zone.h
index afc10a67a777..85e159d8f935 100644
--- a/include/linux/pstore_zone.h
+++ b/include/linux/pstore_zone.h
@@ -16,6 +16,7 @@
* than 4096 and be multiple of 4096.
* @oops_size: The size of oops/panic zone. Zero means disabled, otherwise,
* it must be multiple of SECTOR_SIZE(512 Bytes).
+ * @pmsg_size: The size of pmsg zone which is the same as @oops_size.
* @dump_oops: Whether to dump oops log.
* @read: The general read operation. Both of the function parameters
* @size and @offset are relative value to storage.
@@ -32,6 +33,7 @@ struct psz_info {

unsigned long total_size;
unsigned long oops_size;
+ unsigned long pmsg_size;
int dump_oops;
psz_read_op read;
psz_write_op write;
--
1.9.1