Re: [PATCH v4] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

[Suraj Sonawane <surajsonawane0215@gmail.com>]

On 16/11/24 01:00, Ira Weiny wrote:
> Suraj Sonawane wrote:
>> Fix an issue detected by syzbot with KASAN:
>>
>> BUG: KASAN: vmalloc-out-of-bounds in cmd_to_func drivers/acpi/nfit/
>> core.c:416 [inline]
>> BUG: KASAN: vmalloc-out-of-bounds in acpi_nfit_ctl+0x20e8/0x24a0
>> drivers/acpi/nfit/core.c:459
>>
>> The issue occurs in cmd_to_func when the call_pkg->nd_reserved2
>> array is accessed without verifying that call_pkg points to a buffer
>> that is appropriately sized as a struct nd_cmd_pkg. This can lead
>> to out-of-bounds access and undefined behavior if the buffer does not
>> have sufficient space.
>>
>> To address this, a check was added in acpi_nfit_ctl() to ensure that
>> buf is not NULL and that buf_len is less than sizeof(*call_pkg)
>> before accessing it. This ensures safe access to the members of
>> call_pkg, including the nd_reserved2 array.
>>
>> Reported-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?extid=7534f060ebda6b8b51b3
>> Tested-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
>> Fixes: ebe9f6f19d80 ("acpi/nfit: Fix bus command validation")
>> Signed-off-by: Suraj Sonawane <surajsonawane0215@gmail.com>
>> ---
>> V1: https://lore.kernel.org/lkml/20241111080429.9861-1-surajsonawane0215@gmail.com/
>> V2: Initialized `out_obj` to `NULL` in `acpi_nfit_ctl()` to prevent
>> potential uninitialized variable usage if condition is true.
>> V3: Changed the condition to if (!buf || buf_len < sizeof(*call_pkg))
>> and updated the Fixes tag to reference the correct commit.
>> V4: Removed the explicit cast to maintain the original code style.
>>
>>   drivers/acpi/nfit/core.c | 9 ++++++++-
>>   1 file changed, 8 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
>> index 5429ec9ef..84d8eef2a 100644
>> --- a/drivers/acpi/nfit/core.c
>> +++ b/drivers/acpi/nfit/core.c
>> @@ -454,8 +454,15 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
>>   	if (cmd_rc)
>>   		*cmd_rc = -EINVAL;
>>   
>> -	if (cmd == ND_CMD_CALL)
>> +	if (cmd == ND_CMD_CALL) {
>> +		if (!buf || buf_len < sizeof(*call_pkg)) {
>> +			rc = -EINVAL;
>> +			goto out;
> 
> This goto is wrong.  This will result in ACPI_FREE() being called on an
> undefined out_obj.
> 
> This should be:
> 			return -EINVAL;
> 
Thanks for catching that. You’re correct about the goto issue. I had 
initialized out_obj to NULL in previous versions but missed it in this 
patch. In v5, I’ve re-initialized out_obj to NULL, to avoid unintended 
access. Here’s the updated patch link: 
https://lore.kernel.org/lkml/20241116114027.19303-1-surajsonawane0215@gmail.com/ 


> Ira
> 
>> +		}
>> +
>>   		call_pkg = buf;
>> +	}
>> +
>>   	func = cmd_to_func(nfit_mem, cmd, call_pkg, &family);
>>   	if (func < 0)
>>   		return func;
>> -- 
>> 2.34.1
>>

Best Regards,
Suraj Sonawane