The ia64_mca_cpe_int_handler() and the ia64_mca_cmc_int_handler() are not safe against nested interrupts. The interrupts are enabled too early. The interrupts should remain blocked while we are inside the lock- protected region. (See ..._history_lock.) The interrupts will be enabled in ia64_mca_log_sal_error_record() unless it is called from the MCA/INIT handlers. Please have a look at this patch. Thanks, Zoltan Menyhart Signed-off-by: Zoltan Menyhart, <Zoltan.Menyhart@xxxxxxxx>
--- linux-2.6.24-old/arch/ia64/kernel/mca.c 2008-02-28 09:32:09.000000000 +0100
+++ linux-2.6.24/arch/ia64/kernel/mca.c 2008-02-28 09:32:09.000000000 +0100
@@ -176,6 +176,14 @@
oops_in_progress = 0;
/*
+ * It is not safe enabling interrupts (nor touching locks) in
+ * MCA/INIT handler context.
+ */
+#define MCA_IRQ_NOTSAFE 0 /* Called from the MCA/INIT handlers */
+#define MCA_IRQ_SAFE 1 /* NOT called from the MCA/INIT handlers */
+
+
+/*
* Push messages into buffer, print them later if not urgent.
*/
void ia64_mca_printk(const char *fmt, ...)
@@ -423,19 +431,26 @@
* and wakes up any processes waiting for error records.
*
* Inputs : sal_info_type (Type of error record MCA/CMC/CPE)
- * FIXME: remove MCA and irq_safe.
+ * irq_safe (MCA_IRQ_NOTSAFE == 0 if called from the MCA/INIT handlers,
+ * MCA_IRQ_SAFE == 1 otherwise)
*/
static void
-ia64_mca_log_sal_error_record(int sal_info_type)
+ia64_mca_log_sal_error_record(int sal_info_type, int irq_safe)
{
u8 *buffer;
sal_log_record_header_t *rh;
u64 size;
- int irq_safe = sal_info_type != SAL_INFO_TYPE_MCA;
#ifdef IA64_MCA_DEBUG_INFO
static const char * const rec_name[] = { "MCA", "INIT", "CMC", "CPE" };
#endif
+ if (irq_safe){
+ /* SAL spec says: SAL_GET_STATE_INFO() and SAL_CLEAR_STATE_INFO()
+ * should be called with interrupts enabled, unless we are in
+ * MCA/INIT handlers.
+ */
+ local_irq_enable();
+ }
size = ia64_log_get(sal_info_type, &buffer, irq_safe);
if (!size)
return;
@@ -512,9 +527,6 @@
IA64_MCA_DEBUG("%s: received interrupt vector = %#x on CPU %d\n",
__FUNCTION__, cpe_irq, smp_processor_id());
- /* SAL spec states this should run w/ interrupts enabled */
- local_irq_enable();
-
spin_lock(&cpe_history_lock);
if (!cpe_poll_enabled && cpe_vector >= 0) {
@@ -553,7 +565,7 @@
spin_unlock(&cpe_history_lock);
out:
/* Get the CPE error record and log it */
- ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CPE);
+ ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CPE, MCA_IRQ_SAFE);
return IRQ_HANDLED;
}
@@ -1245,7 +1257,7 @@
}
/* Get the MCA error record and log it */
- ia64_mca_log_sal_error_record(SAL_INFO_TYPE_MCA);
+ ia64_mca_log_sal_error_record(SAL_INFO_TYPE_MCA, MCA_IRQ_NOTSAFE);
/* MCA error recovery */
recover = (ia64_mca_ucmc_extension
@@ -1322,9 +1334,6 @@
IA64_MCA_DEBUG("%s: received interrupt vector = %#x on CPU %d\n",
__FUNCTION__, cmc_irq, smp_processor_id());
- /* SAL spec states this should run w/ interrupts enabled */
- local_irq_enable();
-
spin_lock(&cmc_history_lock);
if (!cmc_polling_enabled) {
int i, count = 1; /* we know 1 happened now */
@@ -1367,7 +1376,7 @@
spin_unlock(&cmc_history_lock);
out:
/* Get the CMC error record and log it */
- ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CMC);
+ ia64_mca_log_sal_error_record(SAL_INFO_TYPE_CMC, MCA_IRQ_SAFE);
return IRQ_HANDLED;
}
|