On Mon, Sep 13, 2010 at 03:23:05AM -0700, Jan Beulich wrote:
> Just like pkgtemp registers another core of the same package when one
> gets removed, coretemp should register another hyperthread (if
> available) in that situation.
>
> As pointed out in the patch fixing the respective code in pkgtemp, the
> list protectng mutex must be dropped before calling
> coretemp_device_add(), and due to the restructured loop (including an
> explicit return) the "safe" variant of the list iterator isn't needed
> anymore.
>
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxxxx>
> Cc: Rudolf Marek <r.marek@xxxxxxxxxxxx>
>
> ---
> drivers/hwmon/coretemp.c | 22 +++++++++++++++-------
> 1 file changed, 15 insertions(+), 7 deletions(-)
>
> --- linux-2.6.36-rc4/drivers/hwmon/coretemp.c 2010-09-13 08:45:02.000000000 +0200
> +++ 2.6.36-rc4-x86-coretemp-remove-deadlock/drivers/hwmon/coretemp.c 2010-09-03 17:54:14.000000000 +0200
> @@ -482,14 +482,22 @@ exit:
>
> static void coretemp_device_remove(unsigned int cpu)
> {
> - struct pdev_entry *p, *n;
> + struct pdev_entry *p;
> + unsigned int i;
> +
> mutex_lock(&pdev_list_mutex);
> - list_for_each_entry_safe(p, n, &pdev_list, list) {
> - if (p->cpu == cpu) {
> - platform_device_unregister(p->pdev);
> - list_del(&p->list);
> - kfree(p);
> - }
> + list_for_each_entry(p, &pdev_list, list) {
> + if (p->cpu != cpu)
> + continue;
> +
> + platform_device_unregister(p->pdev);
> + list_del(&p->list);
> + mutex_unlock(&pdev_list_mutex);
> + kfree(p);
> + for_each_cpu(i, cpu_sibling_mask(cpu))
> + if (i != cpu && !coretemp_device_add(i))
> + break;
> + return;
> }
> mutex_unlock(&pdev_list_mutex);
> }
>
>
>
I sent a patch to fix this issue before. I'll push it to Linus.
BTW, you patch won't be compiled in UP mode.
Thanks.
-Fenghua
From: Fenghua Yu <fenghua.yu@xxxxxxxxx>
In current coretemp driver, when a CPU in dev_list is hot-removed, although its
HT sibling is still running, its core sensor is gone and not available to user
level application any more.
When a CPU is hot-removed, its core sensor should be still available to upper
level application as long as the hot-removed CPU's HT sibling is still running.
A core sensor is invisible to user level only when all of siblings in a core are
hot-removed.
This patch fixes this issue.
Signed-off-by: Fenghua Yu <fenghua.yu@xxxxxxxxx>
---
drivers/hwmon/coretemp.c | 25 ++++++++++++++++++++++---
1 files changed, 22 insertions(+), 3 deletions(-)
diff --git a/drivers/hwmon/coretemp.c b/drivers/hwmon/coretemp.c
index c070c97..2257cc4 100644
--- a/drivers/hwmon/coretemp.c
+++ b/drivers/hwmon/coretemp.c
@@ -418,7 +418,7 @@ struct pdev_entry {
static LIST_HEAD(pdev_list);
static DEFINE_MUTEX(pdev_list_mutex);
-static int __cpuinit coretemp_device_add(unsigned int cpu)
+static int coretemp_device_add(unsigned int cpu)
{
int err;
struct platform_device *pdev;
@@ -483,15 +483,34 @@ exit:
static void coretemp_device_remove(unsigned int cpu)
{
struct pdev_entry *p, *n;
- mutex_lock(&pdev_list_mutex);
+#ifdef CONFIG_SMP
+ int s;
+#endif
+
list_for_each_entry_safe(p, n, &pdev_list, list) {
if (p->cpu == cpu) {
+ mutex_lock(&pdev_list_mutex);
platform_device_unregister(p->pdev);
list_del(&p->list);
kfree(p);
+ mutex_unlock(&pdev_list_mutex);
+
+#ifdef CONFIG_SMP
+ /*
+ * Add removed CPU's HT sibling to dev_list.
+ * If there is no sibling available, the core sensor
+ * is invisiable to user space any more.
+ */
+ for_each_cpu(s, cpu_sibling_mask(cpu)) {
+ if (s != cpu) {
+ coretemp_device_add(s);
+ break;
+ }
+ }
+#endif
+ return;
}
}
- mutex_unlock(&pdev_list_mutex);
}
static int __cpuinit coretemp_cpu_callback(struct notifier_block *nfb,
--
1.6.0.3
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
