cluster-commits November 2016

cluster-commits@lists.stg.fedorahosted.org

3 participants
5 discussions

gfs2-utils: master - fsck.gfs2: Handle gfs2_readi() errors in check_statfs()
by Andrew Price 14 Nov '16

14 Nov '16

Gitweb: http://git.fedorahosted.org/git/?p=gfs2-utils.git;a=commitdiff;h=dcbde3c14a… Commit: dcbde3c14ae223e36daca420bb3a8f3a269a9398 Parent: 9add3ea518e7c64cbea464571a543d4c8365635c Author: Andrew Price <anprice(a)redhat.com> AuthorDate: Fri Nov 11 13:24:34 2016 +0000 Committer: Andrew Price <anprice(a)redhat.com> CommitterDate: Fri Nov 11 13:28:18 2016 +0000 fsck.gfs2: Handle gfs2_readi() errors in check_statfs() Discovered by static analysis with clang. Related: rhbz#1394235 Signed-off-by: Andrew Price <anprice(a)redhat.com> --- gfs2/fsck/main.c | 9 ++++++--- 1 files changed, 6 insertions(+), 3 deletions(-) diff --git a/gfs2/fsck/main.c b/gfs2/fsck/main.c index c5967bc..51f3e03 100644 --- a/gfs2/fsck/main.c +++ b/gfs2/fsck/main.c @@ -170,9 +170,12 @@ static int check_statfs(struct gfs2_sbd *sdp) /* Read the current statfs values */ count = gfs2_readi(sdp->md.statfs, buf, 0, sdp->md.statfs->i_di.di_size); - if (count == sizeof(struct gfs2_statfs_change)) - gfs2_statfs_change_in(&sc, buf); - + if (count != sizeof(struct gfs2_statfs_change)) { + log_err(_("Failed to read statfs values (%d of %"PRIu64" read)\n"), + count, (uint64_t)sdp->md.statfs->i_di.di_size); + return FSCK_ERROR; + } + gfs2_statfs_change_in(&sc, buf); /* Calculate the real values from the rgrp information */ sdp->blks_total = 0; sdp->blks_alloced = 0;

1 0

cluster: STABLE32 - cman_tool: Improve on the last cman_tool nodes fix
by Christine Caulfield 11 Nov '16

11 Nov '16

Gitweb: http://git.fedorahosted.org/git/?p=cluster.git;a=commitdiff;h=1ac358de4743d… Commit: 1ac358de4743de3bc401eac639ca88252231f3d9 Parent: b5c85f77a6e4dfd4237c12d4002464a429e5b59c Author: Christine Caulfield <ccaulfie(a)redhat.com> AuthorDate: Fri Nov 11 13:00:47 2016 +0000 Committer: Christine Caulfield <ccaulfie(a)redhat.com> CommitterDate: Fri Nov 11 13:00:47 2016 +0000 cman_tool: Improve on the last cman_tool nodes fix clang discoved an unitiialied variable error, and that revealed that other things were wrong. now print 'none' for the qdiskd 'ip address' and avoid uninitialised fields. Signed-off-by: Christine Caulfield <ccaulfie(a)redhat.com> --- cman/cman_tool/main.c | 11 +++++------ 1 files changed, 5 insertions(+), 6 deletions(-) diff --git a/cman/cman_tool/main.c b/cman/cman_tool/main.c index 9aaac26..6946e9e 100644 --- a/cman/cman_tool/main.c +++ b/cman/cman_tool/main.c @@ -168,8 +168,10 @@ static void print_address(char *addr) else saddr = &sin->sin_addr; - inet_ntop(ss->ss_family, saddr, buf, sizeof(buf)); - printf("%s", buf); + if (inet_ntop(ss->ss_family, saddr, buf, sizeof(buf))) + printf("%s", buf); + else + printf("none"); } static char *membership_state(char *buf, int buflen, int node_state) @@ -373,7 +375,7 @@ static void print_node(commandline_t *comline, cman_handle_t h, int *format, str { char member_type; struct tm *jtime; - int numaddrs; + int numaddrs=1; struct cman_node_address addrs[MAX_INTERFACES]; char jstring[1024]; int i,j,k; @@ -437,9 +439,6 @@ static void print_node(commandline_t *comline, cman_handle_t h, int *format, str } } } - else { - printf("0.0.0.0"); - } } if (comline->format_opts && (cman_get_node_extra(h, node->cn_nodeid, &enode) == 0)) {

1 0

cluster: STABLE32 - cman_tool: Don't segfault when taking address of qdevice
by Christine Caulfield 11 Nov '16

11 Nov '16

Gitweb: http://git.fedorahosted.org/git/?p=cluster.git;a=commitdiff;h=b5c85f77a6e4d… Commit: b5c85f77a6e4dfd4237c12d4002464a429e5b59c Parent: 9b8e381474c4e7fcbdb192b516cb5a0dd41b1730 Author: Christine Caulfield <ccaulfie(a)redhat.com> AuthorDate: Fri Oct 14 15:28:42 2016 +0100 Committer: Christine Caulfield <ccaulfie(a)redhat.com> CommitterDate: Fri Oct 14 15:28:42 2016 +0100 cman_tool: Don't segfault when taking address of qdevice If cman_tool nodes -F addr is called when qdevice is active then cman_tool segfaults quite spectacularly when it tries to print the address of the qdevice 'node' Signed-off-by: Christine Caulfield <ccaulfie(a)redhat.com> --- cman/cman_tool/main.c | 24 ++++++++++++++---------- 1 files changed, 14 insertions(+), 10 deletions(-) diff --git a/cman/cman_tool/main.c b/cman/cman_tool/main.c index 5596698..9aaac26 100644 --- a/cman/cman_tool/main.c +++ b/cman/cman_tool/main.c @@ -423,19 +423,23 @@ static void print_node(commandline_t *comline, cman_handle_t h, int *format, str } if (comline->addresses_opt || comline->format_opts) { - if (!cman_get_node_addrs(h, node->cn_nodeid, MAX_INTERFACES, &numaddrs, addrs) && - numaddrs) - { - if (!comline->format_opts) { - printf(" Addresses: "); - for (i = 0; i < numaddrs; i++) - { - print_address(addrs[i].cna_address); - printf(" "); + if (node->cn_nodeid > 0) { + if (!cman_get_node_addrs(h, node->cn_nodeid, MAX_INTERFACES, &numaddrs, addrs) && + numaddrs) { + if (!comline->format_opts) { + printf(" Addresses: "); + for (i = 0; i < numaddrs; i++) + { + print_address(addrs[i].cna_address); + printf(" "); + } + printf("\n"); } - printf("\n"); } } + else { + printf("0.0.0.0"); + } } if (comline->format_opts && (cman_get_node_extra(h, node->cn_nodeid, &enode) == 0)) {

1 0

cluster: RHEL6 - rgmanager: Fix possible segfault in clustat
by Ryan McCabe 08 Nov '16

08 Nov '16

Gitweb: http://git.fedorahosted.org/git/?p=cluster.git;a=commitdiff;h=24e142a12dd3b… Commit: 24e142a12dd3b3976cfd2227f2a1844698f5fbbe Parent: 75a02d64a5f11f9a3f1d84a6606e693a259f8c5b Author: Ryan McCabe <rmccabe(a)redhat.com> AuthorDate: Tue Nov 8 09:31:34 2016 -0500 Committer: Ryan McCabe <rmccabe(a)redhat.com> CommitterDate: Tue Nov 8 09:31:34 2016 -0500 rgmanager: Fix possible segfault in clustat Fix a segfault in clustat that could occur when failing to receive the member list from ccs. Resolves: rhbz#1079207 Signed-off-by: Ryan McCabe <rmccabe(a)redhat.com> --- rgmanager/src/utils/clustat.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-) diff --git a/rgmanager/src/utils/clustat.c b/rgmanager/src/utils/clustat.c index 56e2a15..7688bbd 100644 --- a/rgmanager/src/utils/clustat.c +++ b/rgmanager/src/utils/clustat.c @@ -357,8 +357,10 @@ static cluster_member_list_t *ccs_member_list(void) ccs_disconnect(desc); ret->cml_members = nodes; - qsort(ret->cml_members, ret->cml_count, sizeof(cman_node_t), - member_id_sort); + if (ret->cml_members) { + qsort(ret->cml_members, ret->cml_count, sizeof(cman_node_t), + member_id_sort); + } return ret; }

1 0

cluster: RHEL6 - rgmanager: Fix segfault when quorum is lost
by Ryan McCabe 01 Nov '16

01 Nov '16

Gitweb: http://git.fedorahosted.org/git/?p=cluster.git;a=commitdiff;h=75a02d64a5f11… Commit: 75a02d64a5f11f9a3f1d84a6606e693a259f8c5b Parent: 24e5e65be1c9b8b433edc1723b4e13feba0db8ba Author: Ryan McCabe <rmccabe(a)redhat.com> AuthorDate: Tue Nov 1 11:43:47 2016 -0400 Committer: Ryan McCabe <rmccabe(a)redhat.com> CommitterDate: Tue Nov 1 11:43:47 2016 -0400 rgmanager: Fix segfault when quorum is lost Fix a segfault that could occur following recovery caused by a loss of quorum. Patch from John Ruemker <jruemker(a)redhat.com> Resolves: rhbz#1367594 Signed-off-by: Ryan McCabe <rmccabe(a)redhat.com> --- rgmanager/src/daemons/rg_state.c | 12 ++++++++++++ rgmanager/src/daemons/rg_thread.c | 2 +- 2 files changed, 13 insertions(+), 1 deletions(-) diff --git a/rgmanager/src/daemons/rg_state.c b/rgmanager/src/daemons/rg_state.c index fe5d99d..2527f19 100644 --- a/rgmanager/src/daemons/rg_state.c +++ b/rgmanager/src/daemons/rg_state.c @@ -905,6 +905,8 @@ svc_migrate(const char *svcName, int target) return RG_EINVAL; membership = member_list(); + if (!membership) + return RG_EFAIL; m = memb_id_to_p(membership, target); if (!m) { free_member_list(membership); @@ -1815,6 +1817,8 @@ handle_relocate_req(char *svcName, int orig_request, int preferred_target, /* TODO: simplify this and don't keep alloc/freeing member lists */ allowed_nodes = member_list(); + if (!allowed_nodes) + return RG_EFAIL; m = memb_id_to_p(allowed_nodes, preferred_target); if (m && m->cn_member) { @@ -1849,6 +1853,8 @@ handle_relocate_req(char *svcName, int orig_request, int preferred_target, if (preferred_target > 0) { allowed_nodes = member_list(); + if (!allowed_nodes) + return RG_EFAIL; /* Mark everyone except me and the preferred target DOWN for now If we can't start it on the preferred target, then we'll try @@ -1927,6 +1933,8 @@ handle_relocate_req(char *svcName, int orig_request, int preferred_target, allowed_nodes = member_list(); //count_resource_groups(allowed_nodes); } + if (!allowed_nodes) + return RG_EFAIL; member_list_shuffle(allowed_nodes); if (preferred_target > 0) @@ -2081,6 +2089,8 @@ handle_start_req(char *svcName, int req, int *new_owner) need_check = have_exclusive_resources(); membership = member_list(); + if (!membership) + return RG_EFAIL; /* * When a service request is from a user application (eg, clusvcadm), @@ -2218,6 +2228,8 @@ handle_start_remote_req(char *svcName, int req) need_check = have_exclusive_resources(); membership = member_list(); + if (!membership) + return RG_EFAIL; /* XXX ok, so we need to say "should I start this if I was the only cluster member online */ diff --git a/rgmanager/src/daemons/rg_thread.c b/rgmanager/src/daemons/rg_thread.c index c08d8aa..6f06b6f 100644 --- a/rgmanager/src/daemons/rg_thread.c +++ b/rgmanager/src/daemons/rg_thread.c @@ -219,7 +219,7 @@ resgroup_thread_main(void *arg) while(1) { pthread_mutex_lock(&reslist_mutex); pthread_mutex_lock(&my_queue_mutex); - if ((req = rq_next_request(&my_queue)) == NULL) { + if (!rg_quorate() || (req = rq_next_request(&my_queue)) == NULL) { /* We're done. No more requests. We're about to kill our thread, so exit the loop with the lock held. */

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

cluster-commits November 2016