[v2] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

[PATCH v2] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

Posted by Suraj Sonawane 1 year, 1 month ago

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 sized appropriately as a `struct nd_cmd_pkg`. This
could lead to out-of-bounds access and undefined behavior if the
buffer does not have sufficient space.

To address this issue, a check was added in `acpi_nfit_ctl()` to
ensure that `buf` is not `NULL` and `buf_len` is greater than or
equal to `sizeof(struct nd_cmd_pkg)` before casting `buf` to
`struct nd_cmd_pkg *`. This ensures safe access to the members of
`call_pkg`, including the `nd_reserved2` array.

This change preventing out-of-bounds reads.

Reported-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=7534f060ebda6b8b51b3
Tested-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
Fixes: 2d5404caa8c7 ("Linux 6.12-rc7")
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.

 drivers/acpi/nfit/core.c | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index 5429ec9ef..d0e655a9c 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -439,7 +439,7 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
 {
 	struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
 	struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
-	union acpi_object in_obj, in_buf, *out_obj;
+	union acpi_object in_obj, in_buf, *out_obj = NULL;
 	const struct nd_cmd_desc *desc = NULL;
 	struct device *dev = acpi_desc->dev;
 	struct nd_cmd_pkg *call_pkg = NULL;
@@ -454,8 +454,14 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
 	if (cmd_rc)
 		*cmd_rc = -EINVAL;
 
-	if (cmd == ND_CMD_CALL)
-		call_pkg = buf;
+	if (cmd == ND_CMD_CALL) {
+		if (buf == NULL || buf_len < sizeof(struct nd_cmd_pkg)) {
+			rc = -EINVAL;
+			goto out;
+		}
+		call_pkg = (struct nd_cmd_pkg *)buf;
+	}
+
 	func = cmd_to_func(nfit_mem, cmd, call_pkg, &family);
 	if (func < 0)
 		return func;
-- 
2.34.1

Re: [PATCH v2] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

Posted by Alison Schofield 1 year, 1 month ago

On Tue, Nov 12, 2024 at 10:50:35AM +0530, 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 sized appropriately as a `struct nd_cmd_pkg`. This
> could lead to out-of-bounds access and undefined behavior if the
> buffer does not have sufficient space.
> 
> To address this issue, a check was added in `acpi_nfit_ctl()` to
> ensure that `buf` is not `NULL` and `buf_len` is greater than or
> equal to `sizeof(struct nd_cmd_pkg)` before casting `buf` to
> `struct nd_cmd_pkg *`. This ensures safe access to the members of
> `call_pkg`, including the `nd_reserved2` array.
> 
> This change preventing out-of-bounds reads.
> 
> Reported-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
> Closes: https://syzkaller.appspot.com/bug?extid=7534f060ebda6b8b51b3
> Tested-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
> Fixes: 2d5404caa8c7 ("Linux 6.12-rc7")
> Signed-off-by: Suraj Sonawane <surajsonawane0215@gmail.com>

Suraj,

The fixes tag needs to be where the issue originated, not
where you discovered it (which I'm guessing was using 6.12-rc7).

Here's how I find the tag:

$ git blame drivers/acpi/nfit/core.c | grep call_pkg | grep buf
ebe9f6f19d80d drivers/acpi/nfit/core.c (Dan Williams       2019-02-07 14:56:50 -0800  458) 		call_pkg = buf;

$ git log -1 --pretty=fixes ebe9f6f19d80d
Fixes: ebe9f6f19d80 ("acpi/nfit: Fix bus command validation")

I think ^ should be your Fixes tag. 



snip

>

Re: [PATCH v2] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

Posted by Suraj Sonawane 1 year, 1 month ago

On 13/11/24 10:21, Alison Schofield wrote:
> On Tue, Nov 12, 2024 at 10:50:35AM +0530, 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 sized appropriately as a `struct nd_cmd_pkg`. This
>> could lead to out-of-bounds access and undefined behavior if the
>> buffer does not have sufficient space.
>>
>> To address this issue, a check was added in `acpi_nfit_ctl()` to
>> ensure that `buf` is not `NULL` and `buf_len` is greater than or
>> equal to `sizeof(struct nd_cmd_pkg)` before casting `buf` to
>> `struct nd_cmd_pkg *`. This ensures safe access to the members of
>> `call_pkg`, including the `nd_reserved2` array.
>>
>> This change preventing out-of-bounds reads.
>>
>> Reported-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
>> Closes: https://syzkaller.appspot.com/bug?extid=7534f060ebda6b8b51b3
>> Tested-by: syzbot+7534f060ebda6b8b51b3@syzkaller.appspotmail.com
>> Fixes: 2d5404caa8c7 ("Linux 6.12-rc7")
>> Signed-off-by: Suraj Sonawane <surajsonawane0215@gmail.com>
> 
> Suraj,
> 
> The fixes tag needs to be where the issue originated, not
> where you discovered it (which I'm guessing was using 6.12-rc7).
> 
Thank you for your feedback.

> Here's how I find the tag:
> 
> $ git blame drivers/acpi/nfit/core.c | grep call_pkg | grep buf
> ebe9f6f19d80d drivers/acpi/nfit/core.c (Dan Williams       2019-02-07 14:56:50 -0800  458) 		call_pkg = buf;
> 
> $ git log -1 --pretty=fixes ebe9f6f19d80d

Thank you for this detailed explaination.
> Fixes: ebe9f6f19d80 ("acpi/nfit: Fix bus command validation")
> 
> I think ^ should be your Fixes tag.

Yes, I ran the provided steps to verify the commit ID ebe9f6f19d80d and 
verified it.
> 
> 
> snip
> 
>>

I'll update the Fixes tag in the next version of the patch. 
Additionally, I will re-test with syzbot and submit the revised patch 
shortly.

Thank you for your time and feedback!

Best,
Suraj Sonawane

Re: [PATCH v2] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

Posted by Alison Schofield 1 year, 1 month ago

On Tue, Nov 12, 2024 at 10:50:35AM +0530, 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 sized appropriately as a `struct nd_cmd_pkg`. This
> could lead to out-of-bounds access and undefined behavior if the
> buffer does not have sufficient space.
> 
> To address this issue, a check was added in `acpi_nfit_ctl()` to
> ensure that `buf` is not `NULL` and `buf_len` is greater than or
> equal to `sizeof(struct nd_cmd_pkg)` before casting `buf` to
> `struct nd_cmd_pkg *`. This ensures safe access to the members of
> `call_pkg`, including the `nd_reserved2` array.

That all sounds good! A couple of coding conventions fixups suggested
below -

snip

> @@ -439,7 +439,7 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
>  {
>  	struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
>  	struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
> -	union acpi_object in_obj, in_buf, *out_obj;
> +	union acpi_object in_obj, in_buf, *out_obj = NULL;
>  	const struct nd_cmd_desc *desc = NULL;
>  	struct device *dev = acpi_desc->dev;
>  	struct nd_cmd_pkg *call_pkg = NULL;
> @@ -454,8 +454,14 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
>  	if (cmd_rc)
>  		*cmd_rc = -EINVAL;
>  
> -	if (cmd == ND_CMD_CALL)
> -		call_pkg = buf;
> +	if (cmd == ND_CMD_CALL) {
> +		if (buf == NULL || buf_len < sizeof(struct nd_cmd_pkg)) {

Comparison to NULL and sizeof() usage preferred like this:
	if (!buf || buf_len < sizeof(*call_pkg))


-snip
> 
>

Re: [PATCH v2] acpi: nfit: vmalloc-out-of-bounds Read in acpi_nfit_ctl

Posted by Suraj Sonawane 1 year, 1 month ago

On 13/11/24 10:08, Alison Schofield wrote:
> On Tue, Nov 12, 2024 at 10:50:35AM +0530, 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 sized appropriately as a `struct nd_cmd_pkg`. This
>> could lead to out-of-bounds access and undefined behavior if the
>> buffer does not have sufficient space.
>>
>> To address this issue, a check was added in `acpi_nfit_ctl()` to
>> ensure that `buf` is not `NULL` and `buf_len` is greater than or
>> equal to `sizeof(struct nd_cmd_pkg)` before casting `buf` to
>> `struct nd_cmd_pkg *`. This ensures safe access to the members of
>> `call_pkg`, including the `nd_reserved2` array.
> 
> That all sounds good! A couple of coding conventions fixups suggested
> below -
> 
> snip
> 
>> @@ -439,7 +439,7 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
>>   {
>>   	struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
>>   	struct nfit_mem *nfit_mem = nvdimm_provider_data(nvdimm);
>> -	union acpi_object in_obj, in_buf, *out_obj;
>> +	union acpi_object in_obj, in_buf, *out_obj = NULL;
>>   	const struct nd_cmd_desc *desc = NULL;
>>   	struct device *dev = acpi_desc->dev;
>>   	struct nd_cmd_pkg *call_pkg = NULL;
>> @@ -454,8 +454,14 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm,
>>   	if (cmd_rc)
>>   		*cmd_rc = -EINVAL;
>>   
>> -	if (cmd == ND_CMD_CALL)
>> -		call_pkg = buf;
>> +	if (cmd == ND_CMD_CALL) {
>> +		if (buf == NULL || buf_len < sizeof(struct nd_cmd_pkg)) {
> 
> Comparison to NULL and sizeof() usage preferred like this:
> 	if (!buf || buf_len < sizeof(*call_pkg))
> 
> 
> -snip
>>
>>
Thank you for the feedback and your time.

I appreciate your review and coding convention insights. I have studied 
the suggested change and updated the patch.

I will test the updated patch with syzbot and submit a new version very 
shortly.

Best regards,
Suraj Sonawane