Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1213664 > unrolled thread

[PATCH V2 09/10] kasan: Prevent deadlock in kasan reporting

Started by"Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
First post2015-08-26 10:30 +0200
Last post2015-08-30 15:00 +0200
Articles 3 — 2 participants

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.


Contents

  [PATCH V2 09/10] kasan: Prevent deadlock in kasan reporting "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> - 2015-08-26 10:30 +0200
    Re: [PATCH V2 09/10] kasan: Prevent deadlock in kasan reporting Andrey Ryabinin <ryabinin.a.a@gmail.com> - 2015-08-27 14:10 +0200
      Re: [PATCH V2 09/10] kasan: Prevent deadlock in kasan reporting "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> - 2015-08-30 15:00 +0200

#1213664 — [PATCH V2 09/10] kasan: Prevent deadlock in kasan reporting

From"Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Date2015-08-26 10:30 +0200
Subject[PATCH V2 09/10] kasan: Prevent deadlock in kasan reporting
Message-ID<q1EfN-gF-31@gated-at.bofh.it>
We we end up calling kasan_report in real mode, our shadow mapping
for even spinlock variable will show poisoned. This will result
in us calling kasan_report_error with lock_report spin lock held.
To prevent this disable kasan reporting when we are priting
error w.r.t kasan.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/kasan/report.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/mm/kasan/report.c b/mm/kasan/report.c
index 79fbc5d14bd2..82b41eb83e43 100644
--- a/mm/kasan/report.c
+++ b/mm/kasan/report.c
@@ -185,6 +185,10 @@ void kasan_report_error(struct kasan_access_info *info)
 {
 	unsigned long flags;
 
+	/*
+	 * Make sure we don't end up in loop.
+	 */
+	kasan_disable_current();
 	spin_lock_irqsave(&report_lock, flags);
 	pr_err("================================="
 		"=================================\n");
@@ -194,12 +198,17 @@ void kasan_report_error(struct kasan_access_info *info)
 	pr_err("================================="
 		"=================================\n");
 	spin_unlock_irqrestore(&report_lock, flags);
+	kasan_enable_current();
 }
 
 void kasan_report_user_access(struct kasan_access_info *info)
 {
 	unsigned long flags;
 
+	/*
+	 * Make sure we don't end up in loop.
+	 */
+	kasan_disable_current();
 	spin_lock_irqsave(&report_lock, flags);
 	pr_err("================================="
 		"=================================\n");
@@ -212,6 +221,7 @@ void kasan_report_user_access(struct kasan_access_info *info)
 	pr_err("================================="
 		"=================================\n");
 	spin_unlock_irqrestore(&report_lock, flags);
+	kasan_enable_current();
 }
 
 void kasan_report(unsigned long addr, size_t size,
-- 
2.5.0

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [next] | [standalone]


#1214494

FromAndrey Ryabinin <ryabinin.a.a@gmail.com>
Date2015-08-27 14:10 +0200
Message-ID<q24ae-3Zi-15@gated-at.bofh.it>
In reply to#1213664
2015-08-26 11:26 GMT+03:00 Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>:
> We we end up calling kasan_report in real mode, our shadow mapping
> for even spinlock variable will show poisoned.

Generally I agree with this patch. We should disable reports when we
print report as early
as possible to prevent recursion in case of bug in spinlock or printk etc.

But I don't understand what is the problem that you observing.
How we ended up with shadow poisoned for a valid spinlock struct?
And since  shadow poisoned for some valid memory we should get
enormous amount of false positive reports.

> This will result
> in us calling kasan_report_error with lock_report spin lock held.
> To prevent this disable kasan reporting when we are priting
> error w.r.t kasan.
>
> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
> ---
>  mm/kasan/report.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/mm/kasan/report.c b/mm/kasan/report.c
> index 79fbc5d14bd2..82b41eb83e43 100644
> --- a/mm/kasan/report.c
> +++ b/mm/kasan/report.c
> @@ -185,6 +185,10 @@ void kasan_report_error(struct kasan_access_info *info)
>  {
>         unsigned long flags;
>
> +       /*
> +        * Make sure we don't end up in loop.
> +        */
> +       kasan_disable_current();
>         spin_lock_irqsave(&report_lock, flags);
>         pr_err("================================="
>                 "=================================\n");
> @@ -194,12 +198,17 @@ void kasan_report_error(struct kasan_access_info *info)
>         pr_err("================================="
>                 "=================================\n");
>         spin_unlock_irqrestore(&report_lock, flags);
> +       kasan_enable_current();
>  }
>
>  void kasan_report_user_access(struct kasan_access_info *info)
>  {
>         unsigned long flags;
>
> +       /*
> +        * Make sure we don't end up in loop.
> +        */
> +       kasan_disable_current();
>         spin_lock_irqsave(&report_lock, flags);
>         pr_err("================================="
>                 "=================================\n");
> @@ -212,6 +221,7 @@ void kasan_report_user_access(struct kasan_access_info *info)
>         pr_err("================================="
>                 "=================================\n");
>         spin_unlock_irqrestore(&report_lock, flags);
> +       kasan_enable_current();
>  }
>
>  void kasan_report(unsigned long addr, size_t size,
> --
> 2.5.0
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [next] | [standalone]


#1215919

From"Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Date2015-08-30 15:00 +0200
Message-ID<q3ang-11w-11@gated-at.bofh.it>
In reply to#1214494
Andrey Ryabinin <ryabinin.a.a@gmail.com> writes:

> 2015-08-26 11:26 GMT+03:00 Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>:
>> We we end up calling kasan_report in real mode, our shadow mapping
>> for even spinlock variable will show poisoned.
>
> Generally I agree with this patch. We should disable reports when we
> print report as early
> as possible to prevent recursion in case of bug in spinlock or printk etc.
>
> But I don't understand what is the problem that you observing.
> How we ended up with shadow poisoned for a valid spinlock struct?
> And since  shadow poisoned for some valid memory we should get
> enormous amount of false positive reports.
>

I still haven't fully isolated all the .c files which should not be
kasan instrumented. That means in case of ppc64 i ended up calling
kasan _load/_store in real mode. That will result in failure w.r.t
to the above spin_lock code.

-aneesh

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web