aboutsummaryrefslogtreecommitdiffstats
path: root/arch
diff options
context:
space:
mode:
authorHidetoshi Seto <[seto.hidetoshi@jp.fujitsu.com]>2009-08-06 17:51:58 -0400
committerTony Luck <tony.luck@intel.com>2009-09-14 19:19:24 -0400
commit0cced40e7c58b1105aef3ca446da7b158a18a9a6 (patch)
tree401e554f0e69375833e6ef70651eacb44ed50909 /arch
parent5959906ee9dee602a46e49c868a7e543e050d605 (diff)
[IA64] kdump: Short path to freeze CPUs
Setting monarch_cpu = -1 to let slaves frozen might not work, because there might be slaves being late, not entered the rendezvous yet. Such slaves might be caught in while (monarch_cpu == -1) loop. Use kdump_in_progress instead of monarch_cpus to break INIT rendezvous and let all slaves enter DIE_INIT_SLAVE_LEAVE smoothly. And monarch no longer need to manage rendezvous if once kdump_in_progress is set, catch the monarch in DIE_INIT_MONARCH_ENTER then. Signed-off-by: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com> Cc: Vivek Goyal <vgoyal@redhat.com> Cc: Haren Myneni <hbabu@us.ibm.com> Cc: kexec@lists.infradead.org Acked-by: Fenghua Yu <fenghua.yu@intel.com> Signed-off-by: Tony Luck <tony.luck@intel.com>
Diffstat (limited to 'arch')
-rw-r--r--arch/ia64/kernel/crash.c15
-rw-r--r--arch/ia64/kernel/mca.c15
2 files changed, 19 insertions, 11 deletions
diff --git a/arch/ia64/kernel/crash.c b/arch/ia64/kernel/crash.c
index 0995fdc7b299..6631a9dfafdc 100644
--- a/arch/ia64/kernel/crash.c
+++ b/arch/ia64/kernel/crash.c
@@ -127,14 +127,13 @@ machine_crash_shutdown(struct pt_regs *pt)
127 * If an INIT is asserted here: 127 * If an INIT is asserted here:
128 * - All receivers might be slaves, since some of cpus could already 128 * - All receivers might be slaves, since some of cpus could already
129 * be frozen and INIT might be masked on monarch. In this case, 129 * be frozen and INIT might be masked on monarch. In this case,
130 * all slaves will park in while (monarch_cpu == -1) loop before 130 * all slaves will be frozen soon since kdump_in_progress will let
131 * DIE_INIT_SLAVE_ENTER that for waiting monarch enters. 131 * them into DIE_INIT_SLAVE_LEAVE.
132 * => TBD: freeze all slaves
133 * - One might be a monarch, but INIT rendezvous will fail since 132 * - One might be a monarch, but INIT rendezvous will fail since
134 * at least this cpu already have INIT masked so it never join 133 * at least this cpu already have INIT masked so it never join
135 * to the rendezvous. In this case, all slaves and monarch will 134 * to the rendezvous. In this case, all slaves and monarch will
136 * be frozen after timeout of the INIT rendezvous. 135 * be frozen soon with no wait since the INIT rendezvous is skipped
137 * => TBD: freeze them without waiting timeout 136 * by kdump_in_progress.
138 */ 137 */
139 kdump_smp_send_stop(); 138 kdump_smp_send_stop();
140 /* not all cpu response to IPI, send INIT to freeze them */ 139 /* not all cpu response to IPI, send INIT to freeze them */
@@ -187,6 +186,7 @@ kdump_init_notifier(struct notifier_block *self, unsigned long val, void *data)
187 break; 186 break;
188 /* fall through */ 187 /* fall through */
189 case DIE_INIT_SLAVE_LEAVE: 188 case DIE_INIT_SLAVE_LEAVE:
189 case DIE_INIT_MONARCH_ENTER:
190 case DIE_MCA_RENDZVOUS_LEAVE: 190 case DIE_MCA_RENDZVOUS_LEAVE:
191 unw_init_running(kdump_cpu_freeze, NULL); 191 unw_init_running(kdump_cpu_freeze, NULL);
192 break; 192 break;
@@ -217,7 +217,6 @@ kdump_init_notifier(struct notifier_block *self, unsigned long val, void *data)
217 if (kdump_on_init && (nd->sos->rv_rc != 1)) { 217 if (kdump_on_init && (nd->sos->rv_rc != 1)) {
218 if (atomic_inc_return(&kdump_in_progress) != 1) 218 if (atomic_inc_return(&kdump_in_progress) != 1)
219 kdump_freeze_monarch = 1; 219 kdump_freeze_monarch = 1;
220 *(nd->monarch_cpu) = -1;
221 } 220 }
222 break; 221 break;
223 case DIE_INIT_MONARCH_LEAVE: 222 case DIE_INIT_MONARCH_LEAVE:
@@ -228,10 +227,8 @@ kdump_init_notifier(struct notifier_block *self, unsigned long val, void *data)
228 case DIE_MCA_MONARCH_LEAVE: 227 case DIE_MCA_MONARCH_LEAVE:
229 /* *(nd->data) indicate if MCA is recoverable */ 228 /* *(nd->data) indicate if MCA is recoverable */
230 if (kdump_on_fatal_mca && !(*(nd->data))) { 229 if (kdump_on_fatal_mca && !(*(nd->data))) {
231 if (atomic_inc_return(&kdump_in_progress) == 1) { 230 if (atomic_inc_return(&kdump_in_progress) == 1)
232 *(nd->monarch_cpu) = -1;
233 machine_kdump_on_init(); 231 machine_kdump_on_init();
234 }
235 /* We got fatal MCA while kdump!? No way!! */ 232 /* We got fatal MCA while kdump!? No way!! */
236 } 233 }
237 break; 234 break;
diff --git a/arch/ia64/kernel/mca.c b/arch/ia64/kernel/mca.c
index 7b30d21c5190..d2877a7bfe2e 100644
--- a/arch/ia64/kernel/mca.c
+++ b/arch/ia64/kernel/mca.c
@@ -1682,14 +1682,25 @@ ia64_init_handler(struct pt_regs *regs, struct switch_stack *sw,
1682 1682
1683 if (!sos->monarch) { 1683 if (!sos->monarch) {
1684 ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_INIT; 1684 ia64_mc_info.imi_rendez_checkin[cpu] = IA64_MCA_RENDEZ_CHECKIN_INIT;
1685
1686#ifdef CONFIG_KEXEC
1687 while (monarch_cpu == -1 && !atomic_read(&kdump_in_progress))
1688 udelay(1000);
1689#else
1685 while (monarch_cpu == -1) 1690 while (monarch_cpu == -1)
1686 cpu_relax(); /* spin until monarch enters */ 1691 cpu_relax(); /* spin until monarch enters */
1692#endif
1687 1693
1688 NOTIFY_INIT(DIE_INIT_SLAVE_ENTER, regs, (long)&nd, 1); 1694 NOTIFY_INIT(DIE_INIT_SLAVE_ENTER, regs, (long)&nd, 1);
1689 NOTIFY_INIT(DIE_INIT_SLAVE_PROCESS, regs, (long)&nd, 1); 1695 NOTIFY_INIT(DIE_INIT_SLAVE_PROCESS, regs, (long)&nd, 1);
1690 1696
1697#ifdef CONFIG_KEXEC
1698 while (monarch_cpu != -1 && !atomic_read(&kdump_in_progress))
1699 udelay(1000);
1700#else
1691 while (monarch_cpu != -1) 1701 while (monarch_cpu != -1)
1692 cpu_relax(); /* spin until monarch leaves */ 1702 cpu_relax(); /* spin until monarch leaves */
1703#endif
1693 1704
1694 NOTIFY_INIT(DIE_INIT_SLAVE_LEAVE, regs, (long)&nd, 1); 1705 NOTIFY_INIT(DIE_INIT_SLAVE_LEAVE, regs, (long)&nd, 1);
1695 1706