2013-03-10 22:48:32

by Alex Grad

Subject: [PATCH]->[PATCH v2] kgdb: Removed kmalloc returned value cast

While at it, check kmalloc return value.

Signed-off-by: Alex Grad <[email protected]>
---
arch/powerpc/kernel/kgdb.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/kernel/kgdb.c b/arch/powerpc/kernel/kgdb.c
index 5ca82cd..9e81dd8 100644
--- a/arch/powerpc/kernel/kgdb.c
+++ b/arch/powerpc/kernel/kgdb.c
@@ -159,7 +159,10 @@ static int kgdb_singlestep(struct pt_regs *regs)
if (user_mode(regs))
return 0;

- backup_current_thread_info = (struct thread_info *)kmalloc(sizeof(struct thread_info), GFP_KERNEL);
+ backup_current_thread_info = kmalloc(sizeof(struct thread_info), GFP_KERNEL);
+ if (!backup_current_thread_info)
+ return -ENOMEM;
+
/*
* On Book E and perhaps other processors, singlestep is handled on
* the critical exception stack. This causes current_thread_info()
--
1.7.10.4

2013-03-11 01:34:15

by “tiejun.chen”

Subject: Re: [PATCH]->[PATCH v2] kgdb: Removed kmalloc returned value cast

On 03/11/2013 06:39 AM, Alex Grad wrote:
> While at it, check kmalloc return value.
>
> Signed-off-by: Alex Grad <[email protected]>
> ---
> arch/powerpc/kernel/kgdb.c | 5 ++++-
> 1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/arch/powerpc/kernel/kgdb.c b/arch/powerpc/kernel/kgdb.c
> index 5ca82cd..9e81dd8 100644
> --- a/arch/powerpc/kernel/kgdb.c
> +++ b/arch/powerpc/kernel/kgdb.c
> @@ -159,7 +159,10 @@ static int kgdb_singlestep(struct pt_regs *regs)
> if (user_mode(regs))
> return 0;
>
> - backup_current_thread_info = (struct thread_info *)kmalloc(sizeof(struct thread_info), GFP_KERNEL);
> + backup_current_thread_info = kmalloc(sizeof(struct thread_info), GFP_KERNEL);
> + if (!backup_current_thread_info)
> + return -ENOMEM;

I already send a kgdb patchset in "[v3][PATCH 6/6] powerpc/kgdb: remove copying
the thread_info" to remove these stuff since its unnecessary to copy current
thread_info now.

Tiejun

2013-03-11 10:15:00

by “tiejun.chen”

Subject: Re: [PATCH]->[PATCH v2] kgdb: Removed kmalloc returned value cast

On 03/11/2013 09:33 AM, tiejun.chen wrote:
> On 03/11/2013 06:39 AM, Alex Grad wrote:
>> While at it, check kmalloc return value.
>>
>> Signed-off-by: Alex Grad <[email protected]>
>> ---
>> arch/powerpc/kernel/kgdb.c | 5 ++++-
>> 1 file changed, 4 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/powerpc/kernel/kgdb.c b/arch/powerpc/kernel/kgdb.c
>> index 5ca82cd..9e81dd8 100644
>> --- a/arch/powerpc/kernel/kgdb.c
>> +++ b/arch/powerpc/kernel/kgdb.c
>> @@ -159,7 +159,10 @@ static int kgdb_singlestep(struct pt_regs *regs)
>> if (user_mode(regs))
>> return 0;
>>
>> - backup_current_thread_info = (struct thread_info *)kmalloc(sizeof(struct
>> thread_info), GFP_KERNEL);
>> + backup_current_thread_info = kmalloc(sizeof(struct thread_info),
>> GFP_KERNEL);
>> + if (!backup_current_thread_info)
>> + return -ENOMEM;
>
> I already send a kgdb patchset in "[v3][PATCH 6/6] powerpc/kgdb: remove copying
> the thread_info" to remove these stuff since its unnecessary to copy current
> thread_info now.

I guess that patchset needs to be reviewed with more time and I also think I can
leave this copying to be compatible for other platforms in the future, but we
really should fix this problem I introduced right now. So I'd like to fix this
with DEFINE_PER_CPU.

I send a patch, "powerpc/kgdb: use DEFINE_PER_CPU to allocate kgdb's
thread_info", so please take a look at if that version is fine to every one.

Any comments are appreciated.

BTW, I will update that kgdb patchset with this change to resend.

Tiejun