pstore/zone,blk: Add support for pmsg frontend

Add pmsg support to pstore/blk (through pstore/zone). To enable, pmsg_size
must be greater than 0 and a multiple of 4096.
Signed-off-by: WeiXiong Liao <liaoweixiong@allwinnertech.com>
Link: https://lore.kernel.org/lkml/20200511233229.27745-4-keescook@chromium.org/Co-developed-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Link: https://lore.kernel.org/lkml/20200512171932.222102-1-colin.king@canonical.comCo-developed-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Kees Cook <keescook@chromium.org>

fs/pstore/Kconfig

@@ -224,3 +224,15 @@ config PSTORE_BLK_MAX_REASON
 
 	  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.

fs/pstore/blk.c

@@ -27,6 +27,14 @@ module_param(max_reason, int, 0400);
 MODULE_PARM_DESC(max_reason,
 		 "maximum reason for kmsg dump (default 2: Oops and Panic)");
 
+#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");
+
 /*
  * blkdev - the block device to use for pstore storage
  *
@@ -133,6 +141,7 @@ static int psblk_register_do(struct pstore_device_info *dev)
 	}
 
 	verify_size(kmsg_size, 4096, dev->flags & PSTORE_FLAGS_DMESG);
+	verify_size(pmsg_size, 4096, dev->flags & PSTORE_FLAGS_PMSG);
 #undef verify_size
 
 	pstore_zone_info->total_size = dev->total_size;

fs/pstore/zone.c

@@ -27,12 +27,14 @@
  *
  * @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[];
 };
 
@@ -88,9 +90,11 @@ struct pstore_zone {
  * struct psz_context - all about running state of pstore/zone
  *
  * @kpszs: kmsg dump storage zones
+ * @ppsz: pmsg storage zone
  * @kmsg_max_cnt: max count of @kpszs
  * @kmsg_read_cnt: counter of total read kmsg dumps
  * @kmsg_write_cnt: counter of total kmsg dump writes
+ * @pmsg_read_cnt: counter of total read pmsg zone
  * @oops_counter: counter of oops dumps
  * @panic_counter: counter of panic dumps
  * @recovered: whether finished recovering data from storage
@@ -101,9 +105,11 @@ struct pstore_zone {
  */
 struct psz_context {
 	struct pstore_zone **kpszs;
+	struct pstore_zone *ppsz;
 	unsigned int kmsg_max_cnt;
 	unsigned int kmsg_read_cnt;
 	unsigned int kmsg_write_cnt;
+	unsigned int pmsg_read_cnt;
 	/*
 	 * These counters should be calculated during recovery.
 	 * It records the oops/panic times after crashes rather than boots.
@@ -143,15 +149,20 @@ static inline int buffer_datalen(struct pstore_zone *zone)
 	return atomic_read(&zone->buffer->datalen);
 }
 
+static inline int buffer_start(struct pstore_zone *zone)
+{
+	return atomic_read(&zone->buffer->start);
+}
+
 static inline bool is_on_panic(void)
 {
 	return atomic_read(&pstore_zone_cxt.on_panic);
 }
 
-static ssize_t psz_zone_read(struct pstore_zone *zone, char *buf,
+static ssize_t psz_zone_read_buffer(struct pstore_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;
@@ -160,6 +171,18 @@ static ssize_t psz_zone_read(struct pstore_zone *zone, char *buf,
 	return len;
 }
 
+static int psz_zone_read_oldbuf(struct pstore_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 pstore_zone *zone,
 		enum psz_flush_mode flush_mode, const char *buf,
 		size_t len, unsigned long off)
@@ -415,6 +438,93 @@ static int psz_kmsg_recover(struct psz_context *cxt)
 	return ret;
 }
 
+static int psz_recover_zone(struct psz_context *cxt, struct pstore_zone *zone)
+{
+	struct pstore_zone_info *info = cxt->pstore_zone_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
@@ -431,7 +541,12 @@ static inline int psz_recovery(struct psz_context *cxt)
 		return 0;
 
 	ret = psz_kmsg_recover(cxt);
+	if (ret)
+		goto out;
 
+	ret = psz_recover_zone(cxt, cxt->ppsz);
+
+out:
 	if (unlikely(ret))
 		pr_err("recover failed\n");
 	else {
@@ -446,9 +561,17 @@ static int psz_pstore_open(struct pstore_info *psi)
 	struct psz_context *cxt = psi->data;
 
 	cxt->kmsg_read_cnt = 0;
+	cxt->pmsg_read_cnt = 0;
 	return 0;
 }
 
+static inline bool psz_old_ok(struct pstore_zone *zone)
+{
+	if (zone && zone->oldbuf && atomic_read(&zone->oldbuf->datalen))
+		return true;
+	return false;
+}
+
 static inline bool psz_ok(struct pstore_zone *zone)
 {
 	if (zone && zone->buffer && buffer_datalen(zone))
@@ -473,6 +596,25 @@ static inline int psz_kmsg_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 pstore_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 +624,8 @@ static int psz_pstore_erase(struct pstore_record *record)
 		if (record->id >= cxt->kmsg_max_cnt)
 			return -EINVAL;
 		return psz_kmsg_erase(cxt, cxt->kpszs[record->id], record);
+	case PSTORE_TYPE_PMSG:
+		return psz_record_erase(cxt, cxt->ppsz);
 	default:
 		return -EINVAL;
 	}
@@ -555,6 +699,55 @@ static int notrace psz_kmsg_write(struct psz_context *cxt,
 	return 0;
 }
 
+static int notrace psz_record_write(struct pstore_zone *zone,
+		struct pstore_record *record)
+{
+	size_t start, rem;
+	bool is_full_data = false;
+	char *buf;
+	int cnt;
+
+	if (!zone || !record)
+		return -ENOSPC;
+
+	if (atomic_read(&zone->buffer->datalen) >= zone->buffer_size)
+		is_full_data = true;
+
+	cnt = record->size;
+	buf = record->buf;
+	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;
@@ -566,6 +759,8 @@ static int notrace psz_pstore_write(struct pstore_record *record)
 	switch (record->type) {
 	case PSTORE_TYPE_DMESG:
 		return psz_kmsg_write(cxt, record);
+	case PSTORE_TYPE_PMSG:
+		return psz_record_write(cxt->ppsz, record);
 	default:
 		return -EINVAL;
 	}
@@ -581,6 +776,13 @@ static struct pstore_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;
 }
 
@@ -631,7 +833,7 @@ static ssize_t psz_kmsg_read(struct pstore_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_kmsg_header));
 	if (unlikely(size < 0)) {
 		kfree(record->buf);
@@ -641,6 +843,32 @@ static ssize_t psz_kmsg_read(struct pstore_zone *zone,
 	return size + hlen;
 }
 
+static ssize_t psz_record_read(struct pstore_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;
@@ -665,6 +893,9 @@ static ssize_t psz_pstore_read(struct pstore_record *record)
 		readop = psz_kmsg_read;
 		record->id = cxt->kmsg_read_cnt - 1;
 		break;
+	case PSTORE_TYPE_PMSG:
+		readop = psz_record_read;
+		break;
 	default:
 		goto next_zone;
 	}
@@ -720,6 +951,8 @@ static void psz_free_all_zones(struct psz_context *cxt)
 {
 	if (cxt->kpszs)
 		psz_free_zones(&cxt->kpszs, &cxt->kmsg_max_cnt);
+	if (cxt->ppsz)
+		psz_free_zone(&cxt->ppsz);
 }
 
 static struct pstore_zone *psz_init_zone(enum pstore_type_id type,
@@ -753,8 +986,10 @@ static struct pstore_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;
 
@@ -809,19 +1044,28 @@ static int psz_alloc_zones(struct psz_context *cxt)
 	struct pstore_zone_info *info = cxt->pstore_zone_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);
+		cxt->ppsz = NULL;
+		goto free_out;
+	}
 
-	size = info->total_size;
-	cxt->kpszs = psz_init_zones(PSTORE_TYPE_DMESG, &off, size,
+	cxt->kpszs = psz_init_zones(PSTORE_TYPE_DMESG, &off,
+			info->total_size - off_size,
 			info->kmsg_size, &cxt->kmsg_max_cnt);
 	if (IS_ERR(cxt->kpszs)) {
 		err = PTR_ERR(cxt->kpszs);
 		cxt->kpszs = NULL;
-		goto fail_out;
+		goto free_out;
 	}
 
 	return 0;
-fail_out:
+free_out:
+	psz_free_all_zones(cxt);
 	return err;
 }
 
@@ -844,7 +1088,7 @@ int register_pstore_zone(struct pstore_zone_info *info)
 		return -EINVAL;
 	}
 
-	if (!info->kmsg_size) {
+	if (!info->kmsg_size && !info->pmsg_size) {
 		pr_warn("at least one record size must be non-zero\n");
 		return -EINVAL;
 	}
@@ -866,6 +1110,7 @@ int register_pstore_zone(struct pstore_zone_info *info)
 
 	check_size(total_size, 4096);
 	check_size(kmsg_size, SECTOR_SIZE);
+	check_size(pmsg_size, SECTOR_SIZE);
 
 #undef check_size
 
@@ -891,6 +1136,7 @@ int register_pstore_zone(struct pstore_zone_info *info)
 	pr_debug("register %s with properties:\n", info->name);
 	pr_debug("\ttotal size : %ld Bytes\n", info->total_size);
 	pr_debug("\tkmsg size : %ld Bytes\n", info->kmsg_size);
+	pr_debug("\tpmsg size : %ld Bytes\n", info->pmsg_size);
 
 	err = psz_alloc_zones(cxt);
 	if (err) {
@@ -920,6 +1166,10 @@ int register_pstore_zone(struct pstore_zone_info *info)
 			pr_cont(",panic_write");
 		pr_cont(")");
 	}
+	if (info->pmsg_size) {
+		cxt->pstore.flags |= PSTORE_FLAGS_PMSG;
+		pr_cont(" pmsg");
+	}
 	pr_cont("\n");
 
 	err = pstore_register(&cxt->pstore);

include/linux/pstore_zone.h

@@ -17,6 +17,7 @@ typedef ssize_t (*pstore_zone_write_op)(const char *, size_t, loff_t);
  * @kmsg_size:	The size of oops/panic zone. Zero means disabled, otherwise,
  *		it must be multiple of SECTOR_SIZE(512 Bytes).
  * @max_reason: Maximum kmsg dump reason to store.
+ * @pmsg_size:	The size of pmsg zone which is the same as @kmsg_size.
  * @read:	The general read operation. Both of the function parameters
  *		@size and @offset are relative value to storage.
  *		On success, the number of bytes should be returned, others
@@ -33,6 +34,7 @@ struct pstore_zone_info {
 	unsigned long total_size;
 	unsigned long kmsg_size;
 	int max_reason;
+	unsigned long pmsg_size;
 	pstore_zone_read_op read;
 	pstore_zone_write_op write;
 	pstore_zone_write_op panic_write;