[PATCH] proc/vmcore: fix potential memory leak in vmcore_init()

[PATCH] proc/vmcore: fix potential memory leak in vmcore_init()

[Jianglei Nie]

elfcorehdr_alloc() allocates a memory chunk for elfcorehdr_addr with
kzalloc(). If is_vmcore_usable() returns false, elfcorehdr_addr is a
predefined value. If parse_crash_elf_headers() occurs some error and
returns a negetive value, the elfcorehdr_addr should be released with
elfcorehdr_free().

We can fix by calling elfcorehdr_free() when parse_crash_elf_headers()
fails.

Signed-off-by: Jianglei Nie <niejianglei2021@163.com>
---
 fs/proc/vmcore.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c
index 4eaeb645e759..7e028cd1e59d 100644
--- a/fs/proc/vmcore.c
+++ b/fs/proc/vmcore.c
@@ -1568,6 +1568,7 @@ static int __init vmcore_init(void)
 		return rc;
 	rc = parse_crash_elf_headers();
 	if (rc) {
+		elfcorehdr_free(elfcorehdr_addr);
 		pr_warn("Kdump: vmcore not initialized\n");
 		return rc;
 	}
-- 
2.25.1

Re: [PATCH] proc/vmcore: fix potential memory leak in vmcore_init()

[Baoquan He]

On 06/30/22 at 12:52am, Jianglei Nie wrote:
> elfcorehdr_alloc() allocates a memory chunk for elfcorehdr_addr with
> kzalloc(). If is_vmcore_usable() returns false, elfcorehdr_addr is a
> predefined value. If parse_crash_elf_headers() occurs some error and
> returns a negetive value, the elfcorehdr_addr should be released with
> elfcorehdr_free().
> 
> We can fix by calling elfcorehdr_free() when parse_crash_elf_headers()
> fails.
> 
> Signed-off-by: Jianglei Nie <niejianglei2021@163.com>
> ---
>  fs/proc/vmcore.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c
> index 4eaeb645e759..7e028cd1e59d 100644
> --- a/fs/proc/vmcore.c
> +++ b/fs/proc/vmcore.c
> @@ -1568,6 +1568,7 @@ static int __init vmcore_init(void)
>  		return rc;
>  	rc = parse_crash_elf_headers();
>  	if (rc) {
> +		elfcorehdr_free(elfcorehdr_addr);
>  		pr_warn("Kdump: vmcore not initialized\n");
>  		return rc;

Guess it's found by code inspecting since if vmcore_init() failed, kdump
kernel won't do anyting meaningful and reboot, then nobody notice or
care about this leak.

If so, is this better?

diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c
index 7e028cd1e59d..ea2f44d77786 100644
--- a/fs/proc/vmcore.c
+++ b/fs/proc/vmcore.c
@@ -1568,16 +1568,16 @@ static int __init vmcore_init(void)
 		return rc;
 	rc = parse_crash_elf_headers();
 	if (rc) {
-		elfcorehdr_free(elfcorehdr_addr);
 		pr_warn("Kdump: vmcore not initialized\n");
-		return rc;
+		goto fail;
 	}
-	elfcorehdr_free(elfcorehdr_addr);
 	elfcorehdr_addr = ELFCORE_ADDR_ERR;
 
 	proc_vmcore = proc_create("vmcore", S_IRUSR, NULL, &vmcore_proc_ops);
 	if (proc_vmcore)
 		proc_vmcore->size = vmcore_size;
+fail:
+	elfcorehdr_free(elfcorehdr_addr);
 	return 0;
 }
 fs_initcall(vmcore_init);

>  	}
> -- 
> 2.25.1
> 
> 
> _______________________________________________
> kexec mailing list
> kexec@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/kexec
> 

[PATCH] proc/vmcore: fix potential memory leak in vmcore_init()

[Jianglei Nie]

elfcorehdr_alloc() allocates a memory chunk for elfcorehdr_addr with
kzalloc(). If is_vmcore_usable() returns false, elfcorehdr_addr is a
predefined value. If parse_crash_elf_headers() gets some error and
returns a negetive value, the elfcorehdr_addr should be released with
elfcorehdr_free().

Fix it by calling elfcorehdr_free() when parse_crash_elf_headers() fails.

Acked-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Jianglei Nie <niejianglei2021@163.com>
---
 fs/proc/vmcore.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c
index f2aa86c421f2..163cb266f5fd 100644
--- a/fs/proc/vmcore.c
+++ b/fs/proc/vmcore.c
@@ -1568,15 +1568,17 @@ static int __init vmcore_init(void)
 	rc = parse_crash_elf_headers();
 	if (rc) {
 		pr_warn("Kdump: vmcore not initialized\n");
-		return rc;
+		goto fail;
 	}
-	elfcorehdr_free(elfcorehdr_addr);
 	elfcorehdr_addr = ELFCORE_ADDR_ERR;
 
 	proc_vmcore = proc_create("vmcore", S_IRUSR, NULL, &vmcore_proc_ops);
 	if (proc_vmcore)
 		proc_vmcore->size = vmcore_size;
-	return 0;
+
+fail:
+	elfcorehdr_free(elfcorehdr_addr);
+	return rc;
 }
 fs_initcall(vmcore_init);
 
-- 
2.25.1

Re: [PATCH] proc/vmcore: fix potential memory leak in vmcore_init()

[Matthew Wilcox]

On Tue, Sep 13, 2022 at 02:25:01PM +0800, Jianglei Nie wrote:
>  	}
> -	elfcorehdr_free(elfcorehdr_addr);
>  	elfcorehdr_addr = ELFCORE_ADDR_ERR;
>  
>  	proc_vmcore = proc_create("vmcore", S_IRUSR, NULL, &vmcore_proc_ops);
>  	if (proc_vmcore)
>  		proc_vmcore->size = vmcore_size;
> -	return 0;
> +
> +fail:
> +	elfcorehdr_free(elfcorehdr_addr);
> +	return rc;
>  }

Did you test this?  It looks like you now call
elfcorehdr_free(ELFCORE_ADDR_ERR) if 'rc' is 0.

Re: [PATCH] proc/vmcore: fix potential memory leak in vmcore_init()

[Baoquan He]

On 09/13/22 at 07:35am, Matthew Wilcox wrote:
> On Tue, Sep 13, 2022 at 02:25:01PM +0800, Jianglei Nie wrote:
> >  	}
> > -	elfcorehdr_free(elfcorehdr_addr);
> >  	elfcorehdr_addr = ELFCORE_ADDR_ERR;
> >  
> >  	proc_vmcore = proc_create("vmcore", S_IRUSR, NULL, &vmcore_proc_ops);
> >  	if (proc_vmcore)
> >  		proc_vmcore->size = vmcore_size;
> > -	return 0;
> > +
> > +fail:
> > +	elfcorehdr_free(elfcorehdr_addr);
> > +	return rc;
> >  }
> 
> Did you test this?  It looks like you now call
> elfcorehdr_free(ELFCORE_ADDR_ERR) if 'rc' is 0.

Right, that will cause problem. It's my fault since I suggested
the current change.

Jianglei, please use your v1 change and post again. Sorry for the
incorrect suggestion.