[PATCH 09/10] x86, mce: make mce_log buffer to ring buffer
From: Hidetoshi Seto
Date: Mon Oct 05 2009 - 02:44:31 EST
This patch implements Per-CPU ring buffer data structure.
+ An array is used to hold MCE records. integer "head" indicates
next writing position and integer "tail" indicates next reading
position.
+ To distinguish buffer empty and full, head and tail wrap to 0 at
MCE_LOG_LIMIT instead of MCE_LOG_LEN. Then the real next writing
position is head % MCE_LOG_LEN, and real next reading position is
tail % MCE_LOG_LEN. If buffer is empty, head == tail, if buffer is
full, head % MCE_LOG_LEN == tail % MCE_LOG_LEN and head != tail.
(This piece originates from Huang's patch, titled:
"x86, MCE: Fix bugs and issues of MCE log ring buffer")
Originally-From: Huang Ying <ying.huang@xxxxxxxxx>
Signed-off-by: Hidetoshi Seto <seto.hidetoshi@xxxxxxxxxxxxxx>
---
arch/x86/include/asm/mce.h | 6 +++
arch/x86/kernel/cpu/mcheck/mce.c | 77 +++++++++++++++++++++----------------
2 files changed, 50 insertions(+), 33 deletions(-)
diff --git a/arch/x86/include/asm/mce.h b/arch/x86/include/asm/mce.h
index c5d4144..4b5ef3c 100644
--- a/arch/x86/include/asm/mce.h
+++ b/arch/x86/include/asm/mce.h
@@ -82,6 +82,12 @@ struct mce {
*/
#define MCE_LOG_LEN 32
+#define MCE_LOG_LIMIT (MCE_LOG_LEN * 2 - 1)
+
+static inline int mce_log_index(int n)
+{
+ return n >= MCE_LOG_LEN ? n - MCE_LOG_LEN : n;
+}
struct mce_log_cpu;
diff --git a/arch/x86/kernel/cpu/mcheck/mce.c b/arch/x86/kernel/cpu/mcheck/mce.c
index 655915b..63a7820 100644
--- a/arch/x86/kernel/cpu/mcheck/mce.c
+++ b/arch/x86/kernel/cpu/mcheck/mce.c
@@ -123,7 +123,8 @@ EXPORT_PER_CPU_SYMBOL_GPL(mce_fake_banks);
*/
struct mce_log_cpu {
- unsigned next;
+ int head;
+ int tail;
struct mce entry[MCE_LOG_LEN];
};
@@ -139,32 +140,34 @@ static struct mce_log mcelog = {
void mce_log(struct mce *mce)
{
struct mce_log_cpu *mcelog_cpu = &__get_cpu_var(mce_log_cpus);
- unsigned next, entry;
+ int head, ihead, tail, next;
/* mce->finished must be set to 0 before written to buffer */
mce->finished = 0;
smp_wmb();
do {
- entry = mcelog_cpu->next;
+ head = mcelog_cpu->head;
+ tail = mcelog_cpu->tail;
+ ihead = mce_log_index(head);
+
/*
* When the buffer fills up discard new entries.
* Assume that the earlier errors are the more
- * interesting ones:
+ * interesting.
*/
- if (entry >= MCE_LOG_LEN) {
+ if (ihead == mce_log_index(tail) && head != tail) {
set_bit(MCE_OVERFLOW, (unsigned long *)&mcelog.flags);
return;
}
- smp_rmb();
- next = entry + 1;
- } while (cmpxchg_local(&mcelog_cpu->next, entry, next) != entry);
+ next = head == MCE_LOG_LIMIT ? 0 : head + 1;
+ } while (cmpxchg_local(&mcelog_cpu->head, head, next) != head);
- memcpy(mcelog_cpu->entry + entry, mce, sizeof(struct mce));
+ memcpy(mcelog_cpu->entry + ihead, mce, sizeof(struct mce));
/* ".finished" of MCE record in buffer must be set after copy */
smp_wmb();
- mcelog_cpu->entry[entry].finished = 1;
+ mcelog_cpu->entry[ihead].finished = 1;
/* bit 0 of notify_user should be set after finished be set */
smp_wmb();
@@ -1486,42 +1489,50 @@ static ssize_t mce_read_cpu(int cpu, char __user *inubuf, size_t usize)
{
struct mce_log_cpu *mcelog_cpu = &per_cpu(mce_log_cpus, cpu);
char __user *ubuf = inubuf;
- unsigned prev, next;
- int i, err;
+ int head, tail, pos, i, err = 0;
- next = mcelog_cpu->next;
- if (!next)
+ head = mcelog_cpu->head;
+ tail = mcelog_cpu->tail;
+ if (head == tail)
return 0;
- err = 0;
- prev = 0;
- do {
- for (i = prev; i < next; i++) {
+ for (pos = tail; pos != head && usize >= sizeof(struct mce);
+ pos = pos == MCE_LOG_LIMIT ? 0 : pos+1) {
+ i = mce_log_index(pos);
+ if (!mcelog_cpu->entry[i].finished) {
int timeout = WRITER_TIMEOUT_NS;
while (!mcelog_cpu->entry[i].finished) {
if (timeout-- <= 0) {
memset(mcelog_cpu->entry + i, 0,
sizeof(struct mce));
+ head = mcelog_cpu->head;
printk(KERN_WARNING "mcelog: timeout "
"waiting for writer to finish!\n");
goto timeout;
}
ndelay(1);
}
- smp_rmb();
- err |= copy_to_user(ubuf, mcelog_cpu->entry + i,
- sizeof(struct mce));
- ubuf += sizeof(struct mce);
-timeout:
- ;
}
-
- memset(mcelog_cpu->entry + prev, 0,
- (next - prev) * sizeof(struct mce));
- prev = next;
- next = cmpxchg(&mcelog_cpu->next, prev, 0);
- } while (next != prev);
+ /*
+ * finished field should be checked before
+ * copy_to_user()
+ */
+ smp_rmb();
+ err |= copy_to_user(ubuf, mcelog_cpu->entry + i,
+ sizeof(struct mce));
+ ubuf += sizeof(struct mce);
+ usize -= sizeof(struct mce);
+ mcelog_cpu->entry[i].finished = 0;
+timeout:
+ ;
+ }
+ /*
+ * mcelog_cpu->tail must be updated after ".finished" of
+ * corresponding MCE records are clear.
+ */
+ smp_wmb();
+ mcelog_cpu->tail = pos;
return err ? -EFAULT : ubuf - inubuf;
}
@@ -1533,7 +1544,7 @@ static int mce_empty(void)
for_each_possible_cpu(cpu) {
mcelog_cpu = &per_cpu(mce_log_cpus, cpu);
- if (mcelog_cpu->next)
+ if (mcelog_cpu->head != mcelog_cpu->tail)
return 0;
}
return 1;
@@ -1548,14 +1559,14 @@ static ssize_t mce_read(struct file *filp, char __user *inubuf, size_t usize,
int cpu, err = 0;
/* Only supports full reads right now */
- if (*off != 0 || usize < sizeof(struct mce) * MCE_LOG_LEN)
+ if (*off != 0 || usize < sizeof(struct mce))
return -EINVAL;
mutex_lock(&mce_read_mutex);
while (!mce_empty()) {
for_each_possible_cpu(cpu) {
- if (usize < MCE_LOG_LEN * sizeof(struct mce))
+ if (usize < sizeof(struct mce))
goto out;
err = mce_read_cpu(cpu, ubuf, sizeof(struct mce));
if (err > 0) {
--
1.6.4.3
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/