path: root/0024-RHBZ-565933-checker-timeout.patch

---
 libmultipath/checkers.h              |   15 +--------------
 libmultipath/checkers/emc_clariion.c |    4 ++--
 libmultipath/checkers/hp_sw.c        |   12 ++++++------
 libmultipath/checkers/libsg.c        |    5 +++--
 libmultipath/checkers/libsg.h        |    3 ++-
 libmultipath/checkers/rdac.c         |    9 +++++----
 libmultipath/checkers/readsector0.c  |    2 +-
 libmultipath/checkers/tur.c          |    4 ++--
 libmultipath/config.h                |    1 +
 libmultipath/dict.c                  |   29 +++++++++++++++++++++++++++++
 libmultipath/discovery.c             |   27 +++++++++++++++++++++++++++
 libmultipath/discovery.h             |    1 +
 libmultipath/propsel.c               |   19 +++++++++++++++++--
 multipath.conf.annotated             |    9 +++++++++
 14 files changed, 106 insertions(+), 34 deletions(-)

Index: multipath-tools/libmultipath/checkers.h
===================================================================
--- multipath-tools.orig/libmultipath/checkers.h
+++ multipath-tools/libmultipath/checkers.h
@@ -69,20 +69,6 @@ enum path_check_state {
 
 #define DEFAULT_CHECKER DIRECTIO
 
-/*
- * Overloaded storage response time can be very long.
- * SG_IO timouts after DEF_TIMEOUT milliseconds, and checkers interprets this
- * as a path failure. multipathd then proactively evicts the path from the DM
- * multipath table in this case.
- *
- * This generaly snow balls and ends up in full eviction and IO errors for end
- * users. Bad. This may also cause SCSI bus resets, causing disruption for all
- * local and external storage hardware users.
- * 
- * Provision a long timeout. Longer than any real-world application would cope
- * with.
- */
-#define DEF_TIMEOUT		300000
 #define ASYNC_TIMEOUT_SEC	30
 
 /*
@@ -98,6 +84,7 @@ struct checker {
 	struct list_head node;
 	int fd;
 	int sync;
+	unsigned int timeout;
 	int disable;
 	char name[CHECKER_NAME_LEN];
 	char message[CHECKER_MSG_LEN];       /* comm with callers */
Index: multipath-tools/libmultipath/checkers/emc_clariion.c
===================================================================
--- multipath-tools.orig/libmultipath/checkers/emc_clariion.c
+++ multipath-tools/libmultipath/checkers/emc_clariion.c
@@ -113,7 +113,7 @@ int libcheck_check (struct checker * c)
 	io_hdr.dxferp = sense_buffer;
 	io_hdr.cmdp = inqCmdBlk;
 	io_hdr.sbp = sb;
-	io_hdr.timeout = DEF_TIMEOUT;
+	io_hdr.timeout = c->timeout;
 	io_hdr.pack_id = 0;
 	if (ioctl(c->fd, SG_IO, &io_hdr) < 0) {
 		MSG(c, "emc_clariion_checker: sending query command failed");
@@ -182,7 +182,7 @@ int libcheck_check (struct checker * c)
 		unsigned char buf[4096];
 
 		memset(buf, 0, 4096);
-		ret = sg_read(c->fd, &buf[0], sbb = &sb[0]);
+		ret = sg_read(c->fd, &buf[0], sbb = &sb[0], c->timeout);
 		if (ret == PATH_DOWN) {
 			hexadecimal_to_ascii(ct->wwn, wwnstr);
 
Index: multipath-tools/libmultipath/checkers/hp_sw.c
===================================================================
--- multipath-tools.orig/libmultipath/checkers/hp_sw.c
+++ multipath-tools/libmultipath/checkers/hp_sw.c
@@ -46,7 +46,7 @@ void libcheck_free (struct checker * c)
 
 static int
 do_inq(int sg_fd, int cmddt, int evpd, unsigned int pg_op,
-       void *resp, int mx_resp_len, int noisy)
+       void *resp, int mx_resp_len, int noisy, unsigned int timeout)
 {
 	unsigned char inqCmdBlk[INQUIRY_CMDLEN] =
 		{ INQUIRY_CMD, 0, 0, 0, 0, 0 };
@@ -70,7 +70,7 @@ do_inq(int sg_fd, int cmddt, int evpd, u
 	io_hdr.dxferp = resp;
 	io_hdr.cmdp = inqCmdBlk;
 	io_hdr.sbp = sense_b;
-	io_hdr.timeout = DEF_TIMEOUT;
+	io_hdr.timeout = timeout;
 
 	if (ioctl(sg_fd, SG_IO, &io_hdr) < 0)
 		return 1;
@@ -98,7 +98,7 @@ do_inq(int sg_fd, int cmddt, int evpd, u
 }
 
 static int
-do_tur (int fd)
+do_tur (int fd, unsigned int timeout)
 {
 	unsigned char turCmdBlk[TUR_CMD_LEN] = { 0x00, 0, 0, 0, 0, 0 };
 	struct sg_io_hdr io_hdr;
@@ -111,7 +111,7 @@ do_tur (int fd)
 	io_hdr.dxfer_direction = SG_DXFER_NONE;
 	io_hdr.cmdp = turCmdBlk;
 	io_hdr.sbp = sense_buffer;
-	io_hdr.timeout = DEF_TIMEOUT;
+	io_hdr.timeout = timeout;
 	io_hdr.pack_id = 0;
 
 	if (ioctl(fd, SG_IO, &io_hdr) < 0)
@@ -128,12 +128,12 @@ libcheck_check (struct checker * c)
 {
 	char buff[MX_ALLOC_LEN];
 
-	if (0 != do_inq(c->fd, 0, 1, 0x80, buff, MX_ALLOC_LEN, 0)) {
+	if (0 != do_inq(c->fd, 0, 1, 0x80, buff, MX_ALLOC_LEN, 0, c->timeout)) {
 		MSG(c, MSG_HP_SW_DOWN);
 		return PATH_DOWN;
 	}
 
-	if (do_tur(c->fd)) {
+	if (do_tur(c->fd, c->timeout)) {
 		MSG(c, MSG_HP_SW_GHOST);
 		return PATH_GHOST;
 	}
Index: multipath-tools/libmultipath/checkers/libsg.c
===================================================================
--- multipath-tools.orig/libmultipath/checkers/libsg.c
+++ multipath-tools/libmultipath/checkers/libsg.c
@@ -11,7 +11,8 @@
 #include "../libmultipath/sg_include.h"
 
 int
-sg_read (int sg_fd, unsigned char * buff, unsigned char * senseBuff)
+sg_read (int sg_fd, unsigned char * buff, unsigned char * senseBuff,
+	 unsigned int timeout)
 {
 	/* defaults */
 	int blocks = 1;
@@ -51,7 +52,7 @@ sg_read (int sg_fd, unsigned char * buff
 	io_hdr.dxferp = buff;
 	io_hdr.mx_sb_len = SENSE_BUFF_LEN;
 	io_hdr.sbp = senseBuff;
-	io_hdr.timeout = DEF_TIMEOUT;
+	io_hdr.timeout = timeout;
 	io_hdr.pack_id = (int)start_block;
 	if (diop && *diop)
 	io_hdr.flags |= SG_FLAG_DIRECT_IO;
Index: multipath-tools/libmultipath/checkers/libsg.h
===================================================================
--- multipath-tools.orig/libmultipath/checkers/libsg.h
+++ multipath-tools/libmultipath/checkers/libsg.h
@@ -3,6 +3,7 @@
 
 #define SENSE_BUFF_LEN 32
 
-int sg_read (int sg_fd, unsigned char * buff, unsigned char * senseBuff);
+int sg_read (int sg_fd, unsigned char * buff, unsigned char * senseBuff,
+	     unsigned int timeout);
 
 #endif /* _LIBSG_H */
Index: multipath-tools/libmultipath/checkers/rdac.c
===================================================================
--- multipath-tools.orig/libmultipath/checkers/rdac.c
+++ multipath-tools/libmultipath/checkers/rdac.c
@@ -18,7 +18,6 @@
 #define INQUIRY_CMDLEN		6
 #define INQUIRY_CMD		0x12
 #define SENSE_BUFF_LEN		32
-#define RDAC_DEF_TIMEOUT	60000
 #define SCSI_CHECK_CONDITION	0x2
 #define SCSI_COMMAND_TERMINATED	0x22
 #define SG_ERR_DRIVER_SENSE	0x08
@@ -43,7 +42,8 @@ void libcheck_free (struct checker * c)
 }
 
 static int
-do_inq(int sg_fd, unsigned int pg_op, void *resp, int mx_resp_len)
+do_inq(int sg_fd, unsigned int pg_op, void *resp, int mx_resp_len,
+       unsigned int timeout)
 {
 	unsigned char inqCmdBlk[INQUIRY_CMDLEN] = { INQUIRY_CMD, 1, 0, 0, 0, 0 };
 	unsigned char sense_b[SENSE_BUFF_LEN];
@@ -62,7 +62,7 @@ do_inq(int sg_fd, unsigned int pg_op, vo
 	io_hdr.dxferp = resp;
 	io_hdr.cmdp = inqCmdBlk;
 	io_hdr.sbp = sense_b;
-	io_hdr.timeout = RDAC_DEF_TIMEOUT;
+	io_hdr.timeout = timeout;
 
 	if (ioctl(sg_fd, SG_IO, &io_hdr) < 0)
 		return 1;
@@ -103,7 +103,8 @@ libcheck_check (struct checker * c)
 	struct volume_access_inq inq;
 
 	memset(&inq, 0, sizeof(struct volume_access_inq));
-	if (0 != do_inq(c->fd, 0xC9, &inq, sizeof(struct volume_access_inq))) {
+	if (0 != do_inq(c->fd, 0xC9, &inq, sizeof(struct volume_access_inq),
+			c->timeout)) {
 		MSG(c, MSG_RDAC_DOWN);
 		return PATH_DOWN;
 	} else {
Index: multipath-tools/libmultipath/checkers/readsector0.c
===================================================================
--- multipath-tools.orig/libmultipath/checkers/readsector0.c
+++ multipath-tools/libmultipath/checkers/readsector0.c
@@ -29,7 +29,7 @@ int libcheck_check (struct checker * c)
 	unsigned char sbuf[SENSE_BUFF_LEN];
 	int ret;
 
-	ret = sg_read(c->fd, &buf[0], &sbuf[0]);
+	ret = sg_read(c->fd, &buf[0], &sbuf[0], c->timeout);
 
 	switch (ret)
 	{
Index: multipath-tools/libmultipath/checkers/tur.c
===================================================================
--- multipath-tools.orig/libmultipath/checkers/tur.c
+++ multipath-tools/libmultipath/checkers/tur.c
@@ -63,7 +63,7 @@ retry:
 	io_hdr.dxferp = (unsigned char *)resp_buffer;
 	io_hdr.cmdp = inq_cmd;
 	io_hdr.sbp = sense_buffer;
-	io_hdr.timeout = 60; // IOCTL timeout value.
+	io_hdr.timeout = c->timeout; // IOCTL timeout value.
 
 	if (ioctl(c->fd, SG_IO, &io_hdr) < 0) {
 		condlog(0, "SG_IO ioctl failed: %s", strerror(errno));
@@ -148,7 +148,7 @@ libcheck_check (struct checker * c)
 	io_hdr.dxfer_direction = SG_DXFER_NONE;
 	io_hdr.cmdp = turCmdBlk;
 	io_hdr.sbp = sense_buffer;
-	io_hdr.timeout = DEF_TIMEOUT;
+	io_hdr.timeout = c->timeout;
 	io_hdr.pack_id = 0;
 	if (ioctl(c->fd, SG_IO, &io_hdr) < 0) {
 		MSG(c, MSG_TUR_DOWN);
Index: multipath-tools/libmultipath/config.h
===================================================================
--- multipath-tools.orig/libmultipath/config.h
+++ multipath-tools/libmultipath/config.h
@@ -80,6 +80,7 @@ struct config {
 	int max_fds;
 	int force_reload;
 	int queue_without_daemon;
+	int checker_timeout;
 	int daemon;
 	int flush_on_last_del;
 	int attribute_flags;
Index: multipath-tools/libmultipath/dict.c
===================================================================
--- multipath-tools.orig/libmultipath/dict.c
+++ multipath-tools/libmultipath/dict.c
@@ -396,6 +396,25 @@ def_queue_without_daemon(vector strvec)
 }
 
 static int
+def_checker_timeout_handler(vector strvec)
+{
+	unsigned int checker_timeout;
+	char *buff;
+
+	buff = set_value(strvec);
+	if (!buff)
+		return 1;
+
+	if (sscanf(buff, "%u", &checker_timeout) == 1)
+		conf->checker_timeout = checker_timeout;
+	else
+		conf->checker_timeout = 0;
+
+	free(buff);
+	return 0;
+}
+
+static int
 def_pg_timeout_handler(vector strvec)
 {
 	int pg_timeout;
@@ -2068,6 +2087,15 @@ snprint_def_queue_without_daemon (char *
 }
 
 static int
+snprint_def_checker_timeout (char *buff, int len, void *data)
+{
+	if (!conf->checker_timeout)
+		return 0;
+
+	return snprintf(buff, len, "%u", conf->checker_timeout);
+}
+
+static int
 snprint_def_pg_timeout (char * buff, int len, void * data)
 {
 	if (conf->pg_timeout == DEFAULT_PGTIMEOUT)
@@ -2166,6 +2194,7 @@ init_keywords(void)
 	install_keyword("rr_weight", &def_weight_handler, &snprint_def_rr_weight);
 	install_keyword("no_path_retry", &def_no_path_retry_handler, &snprint_def_no_path_retry);
 	install_keyword("queue_without_daemon", &def_queue_without_daemon, &snprint_def_queue_without_daemon);
+	install_keyword("checker_timeout", &def_checker_timeout_handler, &snprint_def_checker_timeout);
 	install_keyword("pg_timeout", &def_pg_timeout_handler, &snprint_def_pg_timeout);
 	install_keyword("flush_on_last_del", &def_flush_on_last_del_handler, &snprint_def_flush_on_last_del);
 	install_keyword("user_friendly_names", &names_handler, &snprint_def_user_friendly_names);
Index: multipath-tools/libmultipath/discovery.c
===================================================================
--- multipath-tools.orig/libmultipath/discovery.c
+++ multipath-tools/libmultipath/discovery.c
@@ -164,6 +164,31 @@ sysfs_get_dev (struct sysfs_device * dev
 }
 
 int
+sysfs_get_timeout(struct sysfs_device *dev, unsigned int *timeout)
+{
+	char *attr;
+	char attr_path[SYSFS_PATH_SIZE];
+	int r;
+	unsigned int t;
+
+	if (safe_sprintf(attr_path, "%s/device", dev->devpath))
+		return 1;
+
+	attr = sysfs_attr_get_value(dev->devpath, "timeout");
+	if (!attr)
+		return 1;
+
+	r = sscanf(attr, "%u\n", &t);
+
+	if (r != 1)
+		return 1;
+
+	*timeout = t * 1000;
+
+	return 0;
+}
+
+int
 sysfs_get_size (struct sysfs_device * dev, unsigned long long * size)
 {
 	char *attr;
@@ -791,6 +816,8 @@ get_state (struct path * pp, int daemon)
 			return PATH_PENDING;
 		checker_set_async(c);
 	}
+	if (!conf->checker_timeout)
+		sysfs_get_timeout(pp->sysdev, &(c->timeout));
 	state = checker_check(c);
 	condlog(3, "%s: state = %i", pp->dev, state);
 	if (state == PATH_DOWN && strlen(checker_message(c)))
Index: multipath-tools/libmultipath/propsel.c
===================================================================
--- multipath-tools.orig/libmultipath/propsel.c
+++ multipath-tools/libmultipath/propsel.c
@@ -16,6 +16,7 @@
 #include "defaults.h"
 #include "devmapper.h"
 #include "prio.h"
+#include "discovery.h"
 
 pgpolicyfn *pgpolicies[] = {
 	NULL,
@@ -274,17 +275,31 @@ select_checker(struct path *pp)
 		checker_get(c, pp->hwe->checker_name);
 		condlog(3, "%s: path checker = %s (controller setting)",
 			pp->dev, checker_name(c));
-		return 0;
+		goto out;
 	}
 	if (conf->checker_name) {
 		checker_get(c, conf->checker_name);
 		condlog(3, "%s: path checker = %s (config file default)",
 			pp->dev, checker_name(c));
-		return 0;
+		goto out;
 	}
 	checker_get(c, DEFAULT_CHECKER);
 	condlog(3, "%s: path checker = %s (internal default)",
 		pp->dev, checker_name(c));
+out:
+	if (conf->checker_timeout) {
+		c->timeout = conf->checker_timeout * 1000;
+		condlog(3, "%s: checker timeout = %u ms (config file default)",
+				pp->dev, c->timeout);
+	}
+	else if (sysfs_get_timeout(pp->sysdev, &c->timeout) == 0)
+		condlog(3, "%s: checker timeout = %u ms (sysfs setting)",
+				pp->dev, c->timeout);
+	else {
+		c->timeout = DEF_TIMEOUT;
+		condlog(3, "%s: checker timeout = %u ms (internal default)",
+				pp->dev, c->timeout);
+	}
 	return 0;
 }
 
Index: multipath-tools/multipath.conf.annotated
===================================================================
--- multipath-tools.orig/multipath.conf.annotated
+++ multipath-tools/multipath.conf.annotated
@@ -202,6 +202,15 @@
 #	gid disk
 #
 #	#
+#	# name    : checker_timeout
+#	# scope   : multipath & multipathd
+#	# desc    : The timeout to use for path checkers that issue scsi
+#	#           commands with an explicit timeout, in seconds.
+#	# values  : n > 0
+#	# default : taken from /sys/block/sd<x>/device/timeout
+#	checker_timeout 60
+#
+#	#
 #	# name    : fast_io_fail_tmo
 #	# scope   : multipath & multipathd
 #	# desc    : The number of seconds the scsi layer will wait after a
Index: multipath-tools/libmultipath/discovery.h
===================================================================
--- multipath-tools.orig/libmultipath/discovery.h
+++ multipath-tools/libmultipath/discovery.h
@@ -36,6 +36,7 @@ int pathinfo (struct path *, vector hwta
 struct path * store_pathinfo (vector pathvec, vector hwtable,
 			      char * devname, int flag);
 int sysfs_set_scsi_tmo (struct multipath *mpp);
+int sysfs_get_timeout(struct sysfs_device *dev, unsigned int *timeout);
 
 /*
  * discovery bitmask