cluster-commits June 2010

cluster-commits@lists.stg.fedorahosted.org

7 participants
65 discussions

cluster: RHEL6 - resource-agents: fix incorrect link resolution in fs-lib
by Lon Hohberger 30 Jun '10

30 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=56… Commit: 56e843fbbce29a4e821f32ebcd271ac0455a4203 Parent: 1f423c3a2d0de6960955fe17dcc5f61f3a709016 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Wed Jun 30 12:22:00 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Jun 30 12:22:00 2010 -0400 resource-agents: fix incorrect link resolution in fs-lib Fs-lib.sh was not resolving symbolic links prior to checking [ -b ], meaning that we would end up with a potential mismatch if what is in /proc/mounts did not match what was in cluster.conf, causing a service failure Resolves: rhnz#609579 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/utils/fs-lib.sh | 12 ++++++------ 1 files changed, 6 insertions(+), 6 deletions(-) diff --git a/rgmanager/src/resources/utils/fs-lib.sh b/rgmanager/src/resources/utils/fs-lib.sh index bc3877e..32667ed 100644 --- a/rgmanager/src/resources/utils/fs-lib.sh +++ b/rgmanager/src/resources/utils/fs-lib.sh @@ -116,12 +116,6 @@ real_device() [ -z "$dev" ] && return $OCF_ERR_ARGS - # If our provided blockdev is a device, we are done - if [ -b "$dev" ]; then - REAL_DEVICE="$dev" - return $OCF_SUCCESS - fi - # Oops, we have a link. Sorry, this is going to fork. if [ -h "$dev" ]; then realdev=$(readlink -f $dev) @@ -132,6 +126,12 @@ real_device() return $OCF_SUCCESS fi + # If our provided blockdev is a device, we are done + if [ -b "$dev" ]; then + REAL_DEVICE="$dev" + return $OCF_SUCCESS + fi + # It's not a link, it's not a block device. If it also # does not match UUID= or LABEL=, then findfs is not # going to find anything useful, so we should quit now.

1 0

cluster: STABLE3 - resource-agents: fix incorrect link resolution in fs-lib
by Lon Hohberger 30 Jun '10

30 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=4d… Commit: 4d24a40f7d0b19d145d03c365a9505fe0927b886 Parent: dbe5e99513b5c6e26aeb70baea075e16084f84ec Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Wed Jun 30 12:22:00 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Jun 30 12:23:42 2010 -0400 resource-agents: fix incorrect link resolution in fs-lib Fs-lib.sh was not resolving symbolic links prior to checking [ -b ], meaning that we would end up with a potential mismatch if what is in /proc/mounts did not match what was in cluster.conf, causing a service failure Resolves: rhnz#609579 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/utils/fs-lib.sh | 12 ++++++------ 1 files changed, 6 insertions(+), 6 deletions(-) diff --git a/rgmanager/src/resources/utils/fs-lib.sh b/rgmanager/src/resources/utils/fs-lib.sh index bc3877e..32667ed 100644 --- a/rgmanager/src/resources/utils/fs-lib.sh +++ b/rgmanager/src/resources/utils/fs-lib.sh @@ -116,12 +116,6 @@ real_device() [ -z "$dev" ] && return $OCF_ERR_ARGS - # If our provided blockdev is a device, we are done - if [ -b "$dev" ]; then - REAL_DEVICE="$dev" - return $OCF_SUCCESS - fi - # Oops, we have a link. Sorry, this is going to fork. if [ -h "$dev" ]; then realdev=$(readlink -f $dev) @@ -132,6 +126,12 @@ real_device() return $OCF_SUCCESS fi + # If our provided blockdev is a device, we are done + if [ -b "$dev" ]; then + REAL_DEVICE="$dev" + return $OCF_SUCCESS + fi + # It's not a link, it's not a block device. If it also # does not match UUID= or LABEL=, then findfs is not # going to find anything useful, so we should quit now.

1 0

cluster: RHEL6 - rgmanager: fix staged upgrade compatibility
by Lon Hohberger 30 Jun '10

30 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=1f… Commit: 1f423c3a2d0de6960955fe17dcc5f61f3a709016 Parent: 58da98dd5214da252fac892fe53b03676e46c13b Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Wed Jun 30 11:11:43 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Jun 30 12:12:27 2010 -0400 rgmanager: fix staged upgrade compatibility This breaks rolling upgrade compatibility for STABLE3 and should not be ported to that branch. Resolves: rhbz#609550 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/include/platform.h | 3 +++ rgmanager/include/resgroup.h | 22 ++++++++-------------- 2 files changed, 11 insertions(+), 14 deletions(-) diff --git a/rgmanager/include/platform.h b/rgmanager/include/platform.h index 7681b50..c5dc66e 100644 --- a/rgmanager/include/platform.h +++ b/rgmanager/include/platform.h @@ -47,6 +47,9 @@ Configure is gone... #define swab32(x) x=be_swap32(x) #define swab64(x) x=be_swap64(x) +#define lswab16(x) x=le_swap16(x) +#define lswab32(x) x=le_swap32(x) +#define lswab64(x) x=le_swap64(x) #if defined(__sparc__) #define ALIGNED __attribute__((aligned)) diff --git a/rgmanager/include/resgroup.h b/rgmanager/include/resgroup.h index 3afbce2..c639315 100644 --- a/rgmanager/include/resgroup.h +++ b/rgmanager/include/resgroup.h @@ -28,7 +28,7 @@ typedef struct { char rs_name[64]; /**< Service name */ /* 64 */ - uint32_t rs_id; /**< Service ID */ + uint32_t rs_flags; /**< Service ID */ uint32_t rs_magic; /**< Magic ID */ uint32_t rs_owner; /**< Member ID running service. */ uint32_t rs_last_owner; /**< Last member to run the service. */ @@ -38,23 +38,17 @@ typedef struct { restarts */ uint64_t rs_transition; /**< Last service transition time */ /* 96 */ - uint32_t rs_flags; /**< User setted flags */ - /* 100 */ - uint8_t rs_version; /**< State version */ - uint8_t _pad_[3]; - /* 104 */ } rg_state_t; #define swab_rg_state_t(ptr) \ {\ - swab32((ptr)->rs_id);\ - swab32((ptr)->rs_magic);\ - swab32((ptr)->rs_owner);\ - swab32((ptr)->rs_last_owner);\ - swab32((ptr)->rs_state);\ - swab32((ptr)->rs_restarts);\ - swab64((ptr)->rs_transition);\ - swab32((ptr)->rs_flags);\ + lswab32((ptr)->rs_magic);\ + lswab32((ptr)->rs_owner);\ + lswab32((ptr)->rs_last_owner);\ + lswab32((ptr)->rs_state);\ + lswab32((ptr)->rs_restarts);\ + lswab64((ptr)->rs_transition);\ + lswab32((ptr)->rs_flags);\ } #if 0

1 0

cluster: RHEL6 - resource-agents: Make vm.sh use stop/start timeouts
by Lon Hohberger 30 Jun '10

30 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=58… Commit: 58da98dd5214da252fac892fe53b03676e46c13b Parent: d6aaa9cc1e5de3d472c6c9169cf980cf2c0bea34 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Mon Jun 21 12:22:23 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Jun 30 06:57:48 2010 -0400 resource-agents: Make vm.sh use stop/start timeouts Startup timeouts are handled by calling the status_program (if specified) every few seconds until either the timeout is reached or the status program returns a successful result. Because the combined VM boot + application time was the basis for the 5 minute check interval for depth=10 (where the status_program is called), it has been reduced to 1 minute intervals. The startup timeout, since it may now be waiting for services within the VM, has been increased to 5 minutes from 20 seconds to match the previous status check timing functionality. This patch fixes the previous patch's inadequate handling of start timeouts. Resolves: bz606754 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/vm.sh | 77 ++++++++++++++++++++++++++++++++++++++--- 1 files changed, 72 insertions(+), 5 deletions(-) diff --git a/rgmanager/src/resources/vm.sh b/rgmanager/src/resources/vm.sh index 08fd097..b3cbc2e 100644 --- a/rgmanager/src/resources/vm.sh +++ b/rgmanager/src/resources/vm.sh @@ -261,15 +261,15 @@ meta_data() </parameters> <actions> - <action name="start" timeout="20"/> + <action name="start" timeout="300"/> <action name="stop" timeout="120"/> <action name="status" timeout="10" interval="30"/> <action name="monitor" timeout="10" interval="30"/>  - <action name="status" depth="10" timeout="10" interval="300"/> - <action name="monitor" depth="10" timeout="10" interval="300"/> + <action name="status" depth="10" timeout="20" interval="60"/> + <action name="monitor" depth="10" timeout="20" interval="60"/>  @@ -351,6 +351,31 @@ do_xm_start() } +get_timeout() +{ + declare -i default_timeout=60 + declare -i tout=60 + + if [ -n "$OCF_RESKEY_RGMANAGER_meta_timeout" ]; then + tout=$OCF_RESKEY_RGMANAGER_meta_timeout + elif [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then + tout=$OCF_RESKEY_CRM_meta_timeout + fi + + if [ $tout -eq 0 ]; then + echo $default_timeout + return 0 + fi + if [ $tout -lt 0 ]; then + echo $default_timeout + return 0 + fi + + echo $tout + return 0 +} + + # # Start a virtual machine given the parameters from # the environment. @@ -428,7 +453,7 @@ do_xm_stop() # do_virsh_stop() { - declare -i timeout=60 + declare -i timeout=$(get_timeout) declare -i ret=1 declare state @@ -443,7 +468,7 @@ do_virsh_stop() echo virsh $op $OCF_RESKEY_name ... virsh $op $OCF_RESKEY_name - timeout=60 + timeout=$(get_timeout) while [ $timeout -gt 0 ]; do sleep 5 ((timeout -= 5)) @@ -913,6 +938,42 @@ migrate() return $rv } + +wait_start() +{ + declare -i timeout_remaining=$(get_timeout) + declare -i start_time + declare -i end_time + declare -i delta + declare -i sleep_time + + if [ -z "$OCF_RESKEY_status_program" ]; then + return 0 + fi + + while [ $timeout_remaining -gt 0 ]; do + start_time=$(date +%s) + bash -c "$OCF_RESKEY_status_program" + if [ $? -eq 0 ]; then + return 0 + fi + end_time=$(date +%s) + delta=$(((end_time - start_time))) + sleep_time=$(((5 - delta))) + + ((timeout_remaining -= $delta)) + if [ $sleep_time -gt 0 ]; then + sleep $sleep_time + ((timeout_remaining -= $sleep_time)) + fi + done + + ocf_log err "Start of $OCF_RESOURCE_INSTANCE has failed" + ocf_log err "Timeout exceeded while waiting for \"$OCF_RESKEY_status_program\"" + + return 1 +} + # # # @@ -921,6 +982,12 @@ case $1 in start) validate_all || exit $OCF_ERR_ARGS do_start + rv=$? + if [ $rv -ne 0 ]; then + exit $rv + fi + + wait_start exit $? ;; stop)

1 0

cluster: STABLE3 - resource-agents: Make vm.sh use stop/start timeouts
by Lon Hohberger 30 Jun '10

30 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=db… Commit: dbe5e99513b5c6e26aeb70baea075e16084f84ec Parent: b8235f4200def1297703b3da7721fb05821eb7d9 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Mon Jun 21 12:22:23 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Jun 30 06:53:30 2010 -0400 resource-agents: Make vm.sh use stop/start timeouts Startup timeouts are handled by calling the status_program (if specified) every few seconds until either the timeout is reached or the status program returns a successful result. Because the combined VM boot + application time was the basis for the 5 minute check interval for depth=10 (where the status_program is called), it has been reduced to 1 minute intervals. The startup timeout, since it may now be waiting for services within the VM, has been increased to 5 minutes from 20 seconds to match the previous status check timing functionality. This patch fixes the previous patch's inadequate handling of start timeouts. Resolves: bz606754 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/vm.sh | 77 ++++++++++++++++++++++++++++++++++++++--- 1 files changed, 72 insertions(+), 5 deletions(-) diff --git a/rgmanager/src/resources/vm.sh b/rgmanager/src/resources/vm.sh index 08fd097..b3cbc2e 100644 --- a/rgmanager/src/resources/vm.sh +++ b/rgmanager/src/resources/vm.sh @@ -261,15 +261,15 @@ meta_data() </parameters> <actions> - <action name="start" timeout="20"/> + <action name="start" timeout="300"/> <action name="stop" timeout="120"/> <action name="status" timeout="10" interval="30"/> <action name="monitor" timeout="10" interval="30"/>  - <action name="status" depth="10" timeout="10" interval="300"/> - <action name="monitor" depth="10" timeout="10" interval="300"/> + <action name="status" depth="10" timeout="20" interval="60"/> + <action name="monitor" depth="10" timeout="20" interval="60"/>  @@ -351,6 +351,31 @@ do_xm_start() } +get_timeout() +{ + declare -i default_timeout=60 + declare -i tout=60 + + if [ -n "$OCF_RESKEY_RGMANAGER_meta_timeout" ]; then + tout=$OCF_RESKEY_RGMANAGER_meta_timeout + elif [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then + tout=$OCF_RESKEY_CRM_meta_timeout + fi + + if [ $tout -eq 0 ]; then + echo $default_timeout + return 0 + fi + if [ $tout -lt 0 ]; then + echo $default_timeout + return 0 + fi + + echo $tout + return 0 +} + + # # Start a virtual machine given the parameters from # the environment. @@ -428,7 +453,7 @@ do_xm_stop() # do_virsh_stop() { - declare -i timeout=60 + declare -i timeout=$(get_timeout) declare -i ret=1 declare state @@ -443,7 +468,7 @@ do_virsh_stop() echo virsh $op $OCF_RESKEY_name ... virsh $op $OCF_RESKEY_name - timeout=60 + timeout=$(get_timeout) while [ $timeout -gt 0 ]; do sleep 5 ((timeout -= 5)) @@ -913,6 +938,42 @@ migrate() return $rv } + +wait_start() +{ + declare -i timeout_remaining=$(get_timeout) + declare -i start_time + declare -i end_time + declare -i delta + declare -i sleep_time + + if [ -z "$OCF_RESKEY_status_program" ]; then + return 0 + fi + + while [ $timeout_remaining -gt 0 ]; do + start_time=$(date +%s) + bash -c "$OCF_RESKEY_status_program" + if [ $? -eq 0 ]; then + return 0 + fi + end_time=$(date +%s) + delta=$(((end_time - start_time))) + sleep_time=$(((5 - delta))) + + ((timeout_remaining -= $delta)) + if [ $sleep_time -gt 0 ]; then + sleep $sleep_time + ((timeout_remaining -= $sleep_time)) + fi + done + + ocf_log err "Start of $OCF_RESOURCE_INSTANCE has failed" + ocf_log err "Timeout exceeded while waiting for \"$OCF_RESKEY_status_program\"" + + return 1 +} + # # # @@ -921,6 +982,12 @@ case $1 in start) validate_all || exit $OCF_ERR_ARGS do_start + rv=$? + if [ $rv -ne 0 ]; then + exit $rv + fi + + wait_start exit $? ;; stop)

1 0

cluster: STABLE3 - Revert "resource-agents: Make vm.sh use stop/start timeouts"
by Lon Hohberger 30 Jun '10

30 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=b8… Commit: b8235f4200def1297703b3da7721fb05821eb7d9 Parent: 9e5a1dafc93894612bc6436ac867c333b97c6181 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Tue Jun 29 11:09:40 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Jun 30 06:53:30 2010 -0400 Revert "resource-agents: Make vm.sh use stop/start timeouts" This reverts commit c9bbf4f12402c5e38e51e21d38682712c18ab5ee. --- rgmanager/src/resources/vm.sh | 65 +++------------------------------------- 1 files changed, 5 insertions(+), 60 deletions(-) diff --git a/rgmanager/src/resources/vm.sh b/rgmanager/src/resources/vm.sh index dee7564..08fd097 100644 --- a/rgmanager/src/resources/vm.sh +++ b/rgmanager/src/resources/vm.sh @@ -261,15 +261,15 @@ meta_data() </parameters> <actions> - <action name="start" timeout="300"/> + <action name="start" timeout="20"/> <action name="stop" timeout="120"/> <action name="status" timeout="10" interval="30"/> <action name="monitor" timeout="10" interval="30"/>  - <action name="status" depth="10" timeout="20" interval="60"/> - <action name="monitor" depth="10" timeout="20" interval="60"/> + <action name="status" depth="10" timeout="10" interval="300"/> + <action name="monitor" depth="10" timeout="10" interval="300"/>  @@ -351,31 +351,6 @@ do_xm_start() } -get_timeout() -{ - declare -i default_timeout=60 - declare -i tout - - if [ -n "$OCF_RESKEY_RGMANAGER_meta_timeout" ]; then - tout=$OCF_RESKEY_RGMANAGER_meta_timeout - elif [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then - tout=$OCF_RESKEY_CRM_meta_timeout - fi - - if [ "$tout" -eq "0" ]; then - echo $default_timeout - return 0 - fi - if [ $tout -lt 0 ]; then - echo $default_timeout - return 0 - fi - - echo $tout - return 0 -} - - # # Start a virtual machine given the parameters from # the environment. @@ -453,7 +428,7 @@ do_xm_stop() # do_virsh_stop() { - declare -i timeout=$(get_timeout) + declare -i timeout=60 declare -i ret=1 declare state @@ -468,7 +443,7 @@ do_virsh_stop() echo virsh $op $OCF_RESKEY_name ... virsh $op $OCF_RESKEY_name - timeout=$(get_timeout) + timeout=60 while [ $timeout -gt 0 ]; do sleep 5 ((timeout -= 5)) @@ -938,30 +913,6 @@ migrate() return $rv } - -wait_start() -{ - declare -i timeout_remaining=$(get_timeout) - - if [ -z "$OCF_RESKEY_status_program" ]; then - return 0 - fi - - while [ $timeout_remaining -gt 0 ]; do - bash -c "$OCF_RESKEY_status_program" &> /dev/null - if [ $? -eq 0 ]; then - return 0 - fi - sleep 5 - ((timeout_remaining -= 5)) - done - - ocf_log err "Start of $OCF_RESOURCE_INSTANCE has failed" - ocf_log err "Timeout exceeded while waiting for \"$OCF_RESKEY_status_program\"" - - return 1 -} - # # # @@ -970,12 +921,6 @@ case $1 in start) validate_all || exit $OCF_ERR_ARGS do_start - rv=$? - if [ $rv -ne 0 ]; then - exit $rv - fi - - wait_start exit $? ;; stop)

1 0

cluster: RHEL6 - rgmanager: Use sysrq-b to reboot
by Lon Hohberger 29 Jun '10

29 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=d6… Commit: d6aaa9cc1e5de3d472c6c9169cf980cf2c0bea34 Parent: 4d6b610339b27a30deac7d9e29dc4236824b58ad Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Mon Jun 28 16:28:06 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Jun 29 14:29:02 2010 -0400 rgmanager: Use sysrq-b to reboot This resolves an issue where the reboot() syscall does not complete in rebooting the system, delaying recovery if the main rgmanager process dies. Resolves: rhbz#609181 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/daemons/watchdog.c | 20 ++++++++++++++++++++ 1 files changed, 20 insertions(+), 0 deletions(-) diff --git a/rgmanager/src/daemons/watchdog.c b/rgmanager/src/daemons/watchdog.c index 6efd655..dfecab1 100644 --- a/rgmanager/src/daemons/watchdog.c +++ b/rgmanager/src/daemons/watchdog.c @@ -4,6 +4,7 @@ #include <sys/reboot.h> #include <stdlib.h> #include <sys/mman.h> +#include <fcntl.h> #include <signals.h> #include <logging.h> @@ -39,6 +40,24 @@ redirect_signals(void) } +static int +sysrq_reboot(void) +{ + int fd; + + fd = open("/proc/sysrq-trigger", O_WRONLY|O_SYNC); + if (fd < 0) + return fd; + + write(fd, "b\n", 2); + fsync(fd); + fdatasync(fd); + close(fd); + + return 0; +} + + /** return watchdog's pid, or 0 on failure */ @@ -80,6 +99,7 @@ watchdog_init(void) #else logt_print(LOG_CRIT, "Watchdog: Daemon died, rebooting...\n"); sync(); + sysrq_reboot(); reboot(RB_AUTOBOOT); #endif exit(255);

1 0

cluster: STABLE3 - rgmanager: Use sysrq-b to reboot
by Lon Hohberger 29 Jun '10

29 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=9e… Commit: 9e5a1dafc93894612bc6436ac867c333b97c6181 Parent: d76d382f7452d8a6c6e2585d7f4bfe4f446c115c Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Mon Jun 28 16:28:06 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Jun 29 14:27:41 2010 -0400 rgmanager: Use sysrq-b to reboot This resolves an issue where the reboot() syscall does not complete in rebooting the system, delaying recovery if the main rgmanager process dies. Resolves: rhbz#609181 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/daemons/watchdog.c | 20 ++++++++++++++++++++ 1 files changed, 20 insertions(+), 0 deletions(-) diff --git a/rgmanager/src/daemons/watchdog.c b/rgmanager/src/daemons/watchdog.c index 6efd655..dfecab1 100644 --- a/rgmanager/src/daemons/watchdog.c +++ b/rgmanager/src/daemons/watchdog.c @@ -4,6 +4,7 @@ #include <sys/reboot.h> #include <stdlib.h> #include <sys/mman.h> +#include <fcntl.h> #include <signals.h> #include <logging.h> @@ -39,6 +40,24 @@ redirect_signals(void) } +static int +sysrq_reboot(void) +{ + int fd; + + fd = open("/proc/sysrq-trigger", O_WRONLY|O_SYNC); + if (fd < 0) + return fd; + + write(fd, "b\n", 2); + fsync(fd); + fdatasync(fd); + close(fd); + + return 0; +} + + /** return watchdog's pid, or 0 on failure */ @@ -80,6 +99,7 @@ watchdog_init(void) #else logt_print(LOG_CRIT, "Watchdog: Daemon died, rebooting...\n"); sync(); + sysrq_reboot(); reboot(RB_AUTOBOOT); #endif exit(255);

1 0

cluster: RHEL56 - GFS: genesis w/flocks on NFS over GFS causes corruption
by Bob Peterson 29 Jun '10

29 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=c0… Commit: c0d08ea227d2c502519abff654b6b2881f7891b6 Parent: 08fd6f0c99e97b74db27643f93f70fd65b686e9f Author: Bob Peterson <rpeterso(a)redhat.com> AuthorDate: Tue Sep 22 16:17:39 2009 -0500 Committer: Bob Peterson <rpeterso(a)redhat.com> CommitterDate: Tue Jun 29 11:47:38 2010 -0500 GFS: genesis w/flocks on NFS over GFS causes corruption This patch does a few things to avoid the NFS over GFS corruption on small (stuffed) files: 1. kmap_atomic is used in place of kmap in stuffed_readpage and gfs_commit_write. 2. Function stuffed_readpage now flushes the dcache page after the page is updated in both of those functions as well. 3. In gfs_get_name there was a typo in an if clause. 4. In do_write_buf the new size is written into the incore inode if the file size has increased. rhbz#245024 --- gfs-kernel/src/gfs/ops_address.c | 25 +++++++++++++------------ gfs-kernel/src/gfs/ops_export.c | 2 +- gfs-kernel/src/gfs/ops_file.c | 5 +++++ 3 files changed, 19 insertions(+), 13 deletions(-) diff --git a/gfs-kernel/src/gfs/ops_address.c b/gfs-kernel/src/gfs/ops_address.c index c9ce4e5..7f5538d 100644 --- a/gfs-kernel/src/gfs/ops_address.c +++ b/gfs-kernel/src/gfs/ops_address.c @@ -204,20 +204,20 @@ static int stuffed_readpage(struct gfs_inode *ip, struct page *page) { struct buffer_head *dibh; - void *kaddr; + char *kaddr; int error; error = gfs_get_inode_buffer(ip, &dibh); if (!error) { - kaddr = kmap(page); - memcpy((char *)kaddr, + kaddr = (char *)kmap_atomic(page, KM_USER0); + memcpy(kaddr, dibh->b_data + sizeof(struct gfs_dinode), ip->i_di.di_size); - memset((char *)kaddr + ip->i_di.di_size, + memset(kaddr + ip->i_di.di_size, 0, PAGE_CACHE_SIZE - ip->i_di.di_size); - kunmap(page); - + kunmap_atomic(kaddr, KM_USER0); + flush_dcache_page(page); brelse(dibh); SetPageUptodate(page); @@ -379,18 +379,19 @@ gfs_commit_write(struct file *file, struct page *page, gfs_trans_add_bh(ip->i_gl, dibh); - kaddr = kmap(page); + kaddr = kmap_atomic(page, KM_USER0); memcpy(dibh->b_data + sizeof(struct gfs_dinode) + from, (char *)kaddr + from, to - from); - kunmap(page); - - brelse(dibh); + flush_dcache_page(page); + kunmap_atomic(kaddr, KM_USER0); SetPageUptodate(page); - - if (inode->i_size < file_size) + if (inode->i_size < file_size) { i_size_write(inode, file_size); + mark_inode_dirty(inode); + } + brelse(dibh); } else { error = generic_commit_write(file, page, from, to); if (error) diff --git a/gfs-kernel/src/gfs/ops_export.c b/gfs-kernel/src/gfs/ops_export.c index 2e8b5c4..e721c58 100644 --- a/gfs-kernel/src/gfs/ops_export.c +++ b/gfs-kernel/src/gfs/ops_export.c @@ -213,7 +213,7 @@ int gfs_get_name(struct dentry *parent, char *name, gfs_glock_dq_uninit(&gh); - if (!error & !*name) + if (!error && !*name) error = -ENOENT; return error; diff --git a/gfs-kernel/src/gfs/ops_file.c b/gfs-kernel/src/gfs/ops_file.c index 150f328..455debb 100644 --- a/gfs-kernel/src/gfs/ops_file.c +++ b/gfs-kernel/src/gfs/ops_file.c @@ -976,6 +976,7 @@ do_write_buf(struct file *file, size_t s; ssize_t count = 0; int error; + struct inode *inode = file->f_mapping->host; gfs_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &ghs[num_gh]); @@ -1022,6 +1023,10 @@ do_write_buf(struct file *file, } error = 0; + if (inode->i_size < ip->i_di.di_size) { + i_size_write(inode, ip->i_di.di_size); + mark_inode_dirty(inode); + } out_gunlock: gfs_glock_dq_m(num_gh + 1, ghs);

1 0

cluster: RHEL56 - fsck.gfs2: unaligned access on ia64
by Bob Peterson 29 Jun '10

29 Jun '10

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=08… Commit: 08fd6f0c99e97b74db27643f93f70fd65b686e9f Parent: 6468ae5d37c4571c13cadda3d4f4be2b439d6f35 Author: Bob Peterson <rpeterso(a)redhat.com> AuthorDate: Tue Jun 29 10:54:50 2010 -0500 Committer: Bob Peterson <rpeterso(a)redhat.com> CommitterDate: Tue Jun 29 10:54:50 2010 -0500 fsck.gfs2: unaligned access on ia64 This patch ports the kernel version of the latest gfs2_bitfit function back to user space. This is more efficient and avoids the unaligned access errors on ia64. rhbz#608158 --- gfs2/fsck/fs_bits.h | 11 ----- gfs2/fsck/pass5.c | 1 - gfs2/libgfs2/fs_bits.c | 114 +++++++++++++++++++++++++++--------------------- gfs2/libgfs2/libgfs2.h | 4 +- 4 files changed, 66 insertions(+), 64 deletions(-) diff --git a/gfs2/fsck/fs_bits.h b/gfs2/fsck/fs_bits.h index 15d6679..2838785 100644 --- a/gfs2/fsck/fs_bits.h +++ b/gfs2/fsck/fs_bits.h @@ -28,15 +28,4 @@ struct fs_bitmap }; typedef struct fs_bitmap fs_bitmap_t; -/* functions with blk #'s that are buffer relative */ -uint32_t gfs2_bitcount(unsigned char *buffer, unsigned int buflen, - unsigned char state); -uint32_t gfs2_bitfit(unsigned char *buffer, unsigned int buflen, - uint32_t goal, unsigned char old_state); - -/* functions with blk #'s that are rgrp relative */ -uint32_t gfs2_blkalloc_internal(struct rgrp_list *rgd, uint32_t goal, - unsigned char old_state, - unsigned char new_state, int do_it); - #endif /* __FS_BITS_H__ */ diff --git a/gfs2/fsck/pass5.c b/gfs2/fsck/pass5.c index f76b9dd..e483ade 100644 --- a/gfs2/fsck/pass5.c +++ b/gfs2/fsck/pass5.c @@ -20,7 +20,6 @@ #include "libgfs2.h" #include "fsck.h" -#include "fs_bits.h" #include "util.h" static int convert_mark(uint8_t q, uint32_t *count) diff --git a/gfs2/libgfs2/fs_bits.c b/gfs2/libgfs2/fs_bits.c index aae2958..ee9dfa7 100644 --- a/gfs2/libgfs2/fs_bits.c +++ b/gfs2/libgfs2/fs_bits.c @@ -31,6 +31,41 @@ #define ALIGN(x,a) (((x)+(a)-1)&~((a)-1)) /** + * gfs2_bit_search + * @ptr: Pointer to bitmap data + * @mask: Mask to use (normally 0x55555.... but adjusted for search start) + * @state: The state we are searching for + * + * We xor the bitmap data with a patter which is the bitwise opposite + * of what we are looking for, this gives rise to a pattern of ones + * wherever there is a match. Since we have two bits per entry, we + * take this pattern, shift it down by one place and then and it with + * the original. All the even bit positions (0,2,4, etc) then represent + * successful matches, so we mask with 0x55555..... to remove the unwanted + * odd bit positions. + * + * This allows searching of a whole u64 at once (32 blocks) with a + * single test (on 64 bit arches). + */ + +static inline uint64_t gfs2_bit_search(const unsigned long long *ptr, + unsigned long long mask, + uint8_t state) +{ + unsigned long long tmp; + static const unsigned long long search[] = { + [0] = 0xffffffffffffffffULL, + [1] = 0xaaaaaaaaaaaaaaaaULL, + [2] = 0x5555555555555555ULL, + [3] = 0x0000000000000000ULL, + }; + tmp = le64_to_cpu(*ptr) ^ search[state]; + tmp &= (tmp >> 1); + tmp &= mask; + return tmp; +} + +/** * gfs2_bitfit - Find a free block in the bitmaps * @buffer: the buffer that holds the bitmaps * @buflen: the length (in bytes) of the buffer @@ -39,60 +74,39 @@ * * Return: the block number that was allocated */ -uint32_t gfs2_bitfit(unsigned char *buffer, unsigned int buflen, - uint32_t goal, unsigned char old_state) +unsigned long gfs2_bitfit(const unsigned char *buf, const unsigned int len, + unsigned long goal, unsigned char state) { - const uint8_t *byte, *start, *end; - int bit, startbit; - uint32_t g1, g2, misaligned; - unsigned long *plong; - unsigned long lskipval; - - lskipval = (old_state & GFS2_BLKST_USED) ? LBITSKIP00 : LBITSKIP55; - g1 = (goal / GFS2_NBBY); - start = buffer + g1; - byte = start; - end = buffer + buflen; - g2 = ALIGN(g1, sizeof(unsigned long)); - plong = (unsigned long *)(buffer + g2); - startbit = bit = (goal % GFS2_NBBY) * GFS2_BIT_SIZE; - misaligned = g2 - g1; - if (!misaligned) - goto ulong_aligned; -/* parse the bitmap a byte at a time */ -misaligned: - while (byte < end) { - if (((*byte >> bit) & GFS2_BIT_MASK) == old_state) { - return goal + - (((byte - start) * GFS2_NBBY) + - ((bit - startbit) >> 1)); - } - bit += GFS2_BIT_SIZE; - if (bit >= GFS2_NBBY * GFS2_BIT_SIZE) { - bit = 0; - byte++; - misaligned--; - if (!misaligned) { - plong = (unsigned long *)byte; - goto ulong_aligned; - } - } - } - return BFITNOENT; - -/* parse the bitmap a unsigned long at a time */ -ulong_aligned: - while ((unsigned char *)plong < end) { - if (((*plong) & LBITMASK) != lskipval) - break; - plong++; - } - if ((unsigned char *)plong < end) { - byte = (const uint8_t *)plong; - misaligned += sizeof(unsigned long) - 1; - goto misaligned; + unsigned long spoint = (goal << 1) & ((8 * sizeof(unsigned long)) - 1); + const unsigned long long *ptr = ((unsigned long long *)buf) + (goal >> 5); + const unsigned long long *end = (unsigned long long *) + (buf + ALIGN(len, sizeof(unsigned long long))); + unsigned long long tmp; + unsigned long long mask = 0x5555555555555555ULL; + unsigned long bit; + + if (state > 3) + return 0; + + /* Mask off bits we don't care about at the start of the search */ + mask <<= spoint; + tmp = gfs2_bit_search(ptr, mask, state); + ptr++; + while(tmp == 0 && ptr < end) { + tmp = gfs2_bit_search(ptr, 0x5555555555555555ULL, state); + ptr++; } - return BFITNOENT; + /* Mask off any bits which are more than len bytes from the start */ + if (ptr == end && (len & (sizeof(unsigned long long) - 1))) + tmp &= (((unsigned long long)~0) >> + (64 - 8 * (len & (sizeof(unsigned long long) - 1)))); + /* Didn't find anything, so return */ + if (tmp == 0) + return BFITNOENT; + ptr--; + bit = ffsll(tmp); + bit /= 2; /* two bits per entry in the bitmap */ + return (((const unsigned char *)ptr - buf) * GFS2_NBBY) + bit; } /** diff --git a/gfs2/libgfs2/libgfs2.h b/gfs2/libgfs2/libgfs2.h index c8cb958..8b4a135 100644 --- a/gfs2/libgfs2/libgfs2.h +++ b/gfs2/libgfs2/libgfs2.h @@ -385,8 +385,8 @@ void fix_device_geometry(struct gfs2_sbd *sdp); /* functions with blk #'s that are buffer relative */ uint32_t gfs2_bitcount(unsigned char *buffer, unsigned int buflen, unsigned char state); -uint32_t gfs2_bitfit(unsigned char *buffer, unsigned int buflen, - uint32_t goal, unsigned char old_state); +unsigned long gfs2_bitfit(const unsigned char *buf, const unsigned int len, + unsigned long goal, unsigned char state); /* functions with blk #'s that are rgrp relative */ uint32_t gfs2_blkalloc_internal(struct rgrp_list *rgd, uint32_t goal,

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

cluster-commits June 2010