Skip to contentSkip to conversation

ixgbe: Fix out-of-bounds warning in ixgbe_host_interface_command()

PATCH In review Gustavo A. R. Silva 3 participants 7 messages Mar 17, 2021

Subsystems (3)
Networking Network drivers Intel Ethernet
lore
Share & tools
lore

Gustavo A. R. Silva

author
#0 lore raw

PATCH 1/1 ixgbe: Fix out-of-bounds warning in ixgbe_host_interface_command()

Fix the following out-of-bounds warning by replacing the one-element array in an anonymous union with a pointer:

CC [M] drivers/net/ethernet/intel/ixgbe/ixgbe_common.o
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c: In function ?ixgbe_host_interface_command?:
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3729:13: warning: array subscript 1 is above array bounds of ?u32[1]? {aka ?unsigned int[1]?} [-Warray-bounds]

 3729 |   bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
      |   ~~~~~~~~~~^~~~
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3682:7: note: while referencing ?u32arr?
 3682 |   u32 u32arr[1];
      |       ^~~~~~

This helps with the ongoing efforts to globally enable -Warray-bounds.

Notice that, the usual approach to fix these sorts of issues is to replace the one-element array with a flexible-array member. However, flexible arrays should not be used in unions. That, together with the fact that the array notation is not being affected in any ways, is why the pointer approach was chosen in this case.

Link https://github.com/KSPP/linux/issues/109
Signed-off-by Gustavo A. R. Silva <[email protected]>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
index 62ddb452f862..bff3dc1af702 100644
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
 	u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
 	union {
 		struct ixgbe_hic_hdr hdr;
-		u32 u32arr[1];
+		u32 *u32arr;
 	} *bp = buffer;
 	u16 buf_len, dword_len;
 	s32 status;
Signature · 1 line
2.27.0
#1 lore raw
Quoted 31 lines · Gustavo A. R. Silva

Fix the following out-of-bounds warning by replacing the one-element array in an anonymous union with a pointer:

CC [M] drivers/net/ethernet/intel/ixgbe/ixgbe_common.o
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c: In function ?ixgbe_host_interface_command?:
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3729:13: warning: array subscript 1 is above array bounds of ?u32[1]? {aka ?unsigned int[1]?} [-Warray-bounds]

 3729 |   bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
      |   ~~~~~~~~~~^~~~
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3682:7: note: while referencing ?u32arr?
 3682 |   u32 u32arr[1];
      |       ^~~~~~

This helps with the ongoing efforts to globally enable -Warray-bounds.

Notice that, the usual approach to fix these sorts of issues is to replace the one-element array with a flexible-array member. However, flexible arrays should not be used in unions. That, together with the fact that the array notation is not being affected in any ways, is why the pointer approach was chosen in this case.

Link https://github.com/KSPP/linux/issues/109
Signed-off-by Gustavo A. R. Silva <[email protected]>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +0 -0
index 62ddb452f862..bff3dc1af702 100644

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer, · line 3685

Quoted 8 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
        union {
                struct ixgbe_hic_hdr hdr;
-               u32 u32arr[1];
+               u32 *u32arr;
        } *bp = buffer;
        u16 buf_len, dword_len;
        s32 status;

This looks bogus. An array is inline, a pointer points elsewhere - they're not interchangeable.

#2 lore raw

Hi Jann,

Please, see my comments below...

Quoted 28 lines · Jann Horn
Quoted 26 lines · Gustavo A. R. Silva

Fix the following out-of-bounds warning by replacing the one-element array in an anonymous union with a pointer:

CC [M] drivers/net/ethernet/intel/ixgbe/ixgbe_common.o
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c: In function ?ixgbe_host_interface_command?:
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3729:13: warning: array subscript 1 is above array bounds of ?u32[1]? {aka ?unsigned int[1]?} [-Warray-bounds]

 3729 |   bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
      |   ~~~~~~~~~~^~~~
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3682:7: note: while referencing ?u32arr?
 3682 |   u32 u32arr[1];
      |       ^~~~~~

This helps with the ongoing efforts to globally enable -Warray-bounds.

Notice that, the usual approach to fix these sorts of issues is to replace the one-element array with a flexible-array member. However, flexible arrays should not be used in unions. That, together with the fact that the array notation is not being affected in any ways, is why the pointer approach was chosen in this case.

Link https://github.com/KSPP/linux/issues/109
Signed-off-by Gustavo A. R. Silva <[email protected]>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 16 lines

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer, · line 3685

Quoted 13 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
index 62ddb452f862..bff3dc1af702 100644
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
        union {
                struct ixgbe_hic_hdr hdr;
-               u32 u32arr[1];
+               u32 *u32arr;
        } *bp = buffer;
        u16 buf_len, dword_len;
        s32 status;

This looks bogus. An array is inline, a pointer points elsewhere - they're not interchangeable.

Yep; but in this case these are the only places in the code where _u32arr_ is being used:

3707 /* first pull in the header so we know the buffer length */
3708 for (bi = 0; bi < dword_len; bi++) {
3709 bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
3710 le32_to_cpus(&bp->u32arr[bi]);
3711 }

3727 /* Pull in the rest of the buffer (bi is where we left off) */
3728 for (; bi <= dword_len; bi++) {
3729 bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
3730 le32_to_cpus(&bp->u32arr[bi]);
3731 }

I think it is safe to turn _u32arra_ into a pointer and continue using the array notation in this particular case.

I also mention this in the changelog text:

"Notice that, the usual approach to fix these sorts of issues is to replace the one-element array with a flexible-array member. However, flexible arrays should not be used in unions. That, together with the fact that the array notation is not being affected in any ways, is why the pointer approach was chosen in this case."

Do you see any particular problem with this in the current code?

Another solution for this would be as follows:

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +12 -5
index 62ddb452f862..3ad95281d790 100644
@@ -3677,10 +3677,11 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
                                 bool return_data)
 {
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
-       union {
-               struct ixgbe_hic_hdr hdr;
-               u32 u32arr[1];
-       } *bp = buffer;
+       struct ixgbe_hic_hdr *bp_hdr = buffer;
+       struct {
+               size_t len;
+               u32 u32arr[];
+       } *bp;
        u16 buf_len, dword_len;
        s32 status;
        u32 bi;

@@ -3704,6 +3705,9 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        /* Calculate length in DWORDs */
        dword_len = hdr_size >> 2;
 
+       bp = kmalloc(struct_size(bp, u32arr, dword_len), GFP_KERNEL);
+       bp->len = dword_len;
+       memcpy(bp->u32arr, buffer, flex_array_size(bp, u32arr, bp->len));
        /* first pull in the header so we know the buffer length */
        for (bi = 0; bi < dword_len; bi++) {
                bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);

@@ -3711,7 +3715,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        }
 
        /* If there is any thing in data position pull it in */
-       buf_len = bp->hdr.buf_len;
+       buf_len = bp_hdr->buf_len;
        if (!buf_len)
                goto rel_out;
 

@@ -3724,6 +3728,9 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        /* Calculate length in DWORDs, add 3 for odd lengths */
        dword_len = (buf_len + 3) >> 2;
 
+       bp = krealloc(bp, struct_size(bp, u32arr, dword_len), GFP_KERNEL);
+       bp->len = dword_len;
+       memcpy(&bp->u32arr[bi], ((u32 *)buffer + bi), flex_array_size(bp, u32arr, bp->len-bi));
        /* Pull in the rest of the buffer (bi is where we left off) */
        for (; bi <= dword_len; bi++) {

		^^^^^^

I just noticed it seems there is a bug right there. I think it should be bi < dword_len, instead

                bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);

What do you guys think?

Thanks!
--
Gustavo

#3 lore raw
Quoted 46 lines · Gustavo A. R. Silva
Quoted 28 lines · Jann Horn
Quoted 26 lines · Gustavo A. R. Silva

Fix the following out-of-bounds warning by replacing the one-element array in an anonymous union with a pointer:

CC [M] drivers/net/ethernet/intel/ixgbe/ixgbe_common.o
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c: In function ?ixgbe_host_interface_command?:
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3729:13: warning: array subscript 1 is above array bounds of ?u32[1]? {aka ?unsigned int[1]?} [-Warray-bounds]

 3729 |   bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
      |   ~~~~~~~~~~^~~~
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c:3682:7: note: while referencing ?u32arr?
 3682 |   u32 u32arr[1];
      |       ^~~~~~

This helps with the ongoing efforts to globally enable -Warray-bounds.

Notice that, the usual approach to fix these sorts of issues is to replace the one-element array with a flexible-array member. However, flexible arrays should not be used in unions. That, together with the fact that the array notation is not being affected in any ways, is why the pointer approach was chosen in this case.

Link https://github.com/KSPP/linux/issues/109
Signed-off-by Gustavo A. R. Silva <[email protected]>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 16 lines

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer, · line 3685

Quoted 13 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
index 62ddb452f862..bff3dc1af702 100644
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
        union {
                struct ixgbe_hic_hdr hdr;
-               u32 u32arr[1];
+               u32 *u32arr;
        } *bp = buffer;
        u16 buf_len, dword_len;
        s32 status;

This looks bogus. An array is inline, a pointer points elsewhere - they're not interchangeable.

Quoted 8 lines

Yep; but in this case these are the only places in the code where _u32arr_ is being used:

3707 /* first pull in the header so we know the buffer length */
3708 for (bi = 0; bi < dword_len; bi++) {
3709 bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
3710 le32_to_cpus(&bp->u32arr[bi]);
3711 }

So now line 3709 means: Read a pointer from bp->u32arr (the value being read from there is not actually a valid pointer), and write to that pointer at offset bi. I don't see how that line could execute without crashing.

#4 lore raw
Quoted 26 lines · Jann Horn
Quoted 17 lines
Quoted 8 lines
Quoted 8 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +0 -0
index 62ddb452f862..bff3dc1af702 100644
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
        union {
                struct ixgbe_hic_hdr hdr;
Quoted 8 lines

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer, · line 3686

Quoted 5 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
-               u32 u32arr[1];
+               u32 *u32arr;
        } *bp = buffer;
        u16 buf_len, dword_len;
        s32 status;

This looks bogus. An array is inline, a pointer points elsewhere - they're not interchangeable.

Quoted 8 lines

Yep; but in this case these are the only places in the code where _u32arr_ is being used:

3707 /* first pull in the header so we know the buffer length */
3708 for (bi = 0; bi < dword_len; bi++) {
3709 bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
3710 le32_to_cpus(&bp->u32arr[bi]);
3711 }

Quoted 4 lines

So now line 3709 means: Read a pointer from bp->u32arr (the value being read from there is not actually a valid pointer), and write to that pointer at offset bi. I don't see how that line could execute without crashing.

Yeah; you're right. I see my confusion now. Apparently, there is no escape from allocating heap memory to fix this issue, as I was proposing in my last email.

I really appreciate the feedback. Thanks!
--
Gustavo

#6 lore raw
In reply to Jann Horn
Quoted 34 lines · Jann Horn
Quoted 32 lines · Gustavo A. R. Silva
Quoted 30 lines · Jann Horn
Quoted 17 lines
Quoted 8 lines
Quoted 8 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +0 -0
index 62ddb452f862..bff3dc1af702 100644
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
        union {
                struct ixgbe_hic_hdr hdr;
Quoted 8 lines

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer, · line 3685

Quoted 5 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
-               u32 u32arr[1];
+               u32 *u32arr;
        } *bp = buffer;
        u16 buf_len, dword_len;
        s32 status;

This looks bogus. An array is inline, a pointer points elsewhere - they're not interchangeable.

Quoted 8 lines

Yep; but in this case these are the only places in the code where _u32arr_ is being used:

3707 /* first pull in the header so we know the buffer length */
3708 for (bi = 0; bi < dword_len; bi++) {
3709 bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
3710 le32_to_cpus(&bp->u32arr[bi]);
3711 }

So now line 3709 means: Read a pointer from bp->u32arr (the value being read from there is not actually a valid pointer), and write to that pointer at offset bi. I don't see how that line could execute without crashing.

Quoted 5 lines
Quoted 3 lines

Yeah; you're right. I see my confusion now. Apparently, there is no escape from allocating heap memory to fix this issue, as I was proposing in my last email.

Why? Can't you do something like this?

Yep; it seems you're right. I was thinking in terms of a flexible array. Also, I think I needed more coffee in my system this morning and I need to stop working after midnight. :)

I'll send a proper patch for this, shortly. I'll add your Proposed-by and Co-developed-by tags to the changelog text.

Thanks a lot for the feedback. I really appreciate it. :) --
Gustavo

Quoted 51 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +2 -4
index 62ddb452f862..768fa124105b 100644
@@ -3677,10 +3677,8 @@ s32 ixgbe_host_interface_command(struct
 ixgbe_hw *hw, void *buffer,
                                 bool return_data)
 {
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
-       union {
-               struct ixgbe_hic_hdr hdr;
-               u32 u32arr[1];
-       } *bp = buffer;
+       u32 *bp = buffer;
+       struct ixgbe_hic_hdr hdr;
        u16 buf_len, dword_len;
        s32 status;

        u32 bi;
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +6 -5
@@ -3706,12 +3704,13 @@ s32 ixgbe_host_interface_command(struct
 ixgbe_hw *hw, void *buffer,
 
        /* first pull in the header so we know the buffer length */
        for (bi = 0; bi < dword_len; bi++) {
-               bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
-               le32_to_cpus(&bp->u32arr[bi]);
+               bp[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
+               le32_to_cpus(&bp[bi]);
        }
 
        /* If there is any thing in data position pull it in */
-       buf_len = bp->hdr.buf_len;
+       memcpy(&hdr, bp, sizeof(hdr));
+       buf_len = hdr.buf_len;
        if (!buf_len)
                goto rel_out;

@@ -3726,8 +3725,8 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw
 *hw, void *buffer,
 
        /* Pull in the rest of the buffer (bi is where we left off) */
        for (; bi <= dword_len; bi++) {
-               bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
-               le32_to_cpus(&bp->u32arr[bi]);
+               bp[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
+               le32_to_cpus(&bp[bi]);
        }

rel_out:

#5 lore raw
In reply to Gustavo A. R. Silva
Quoted 27 lines · Gustavo A. R. Silva
Quoted 26 lines · Jann Horn
Quoted 25 lines
Quoted 8 lines
Quoted 8 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +0 -0
index 62ddb452f862..bff3dc1af702 100644
@@ -3679,7 +3679,7 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
        union {
                struct ixgbe_hic_hdr hdr;
Quoted 8 lines

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer, · line 3685

Quoted 5 lines
drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +1 -1
-               u32 u32arr[1];
+               u32 *u32arr;
        } *bp = buffer;
        u16 buf_len, dword_len;
        s32 status;

This looks bogus. An array is inline, a pointer points elsewhere - they're not interchangeable.

Yep; but in this case these are the only places in the code where _u32arr_ is being used:

3707 /* first pull in the header so we know the buffer length */
3708 for (bi = 0; bi < dword_len; bi++) {
3709 bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
3710 le32_to_cpus(&bp->u32arr[bi]);
3711 }

Quoted 8 lines
Quoted 4 lines

So now line 3709 means: Read a pointer from bp->u32arr (the value being read from there is not actually a valid pointer), and write to that pointer at offset bi. I don't see how that line could execute without crashing.

Yeah; you're right. I see my confusion now. Apparently, there is no escape from allocating heap memory to fix this issue, as I was proposing in my last email.

Why? Can't you do something like this?

(quoted diff) +0 -0

b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
index 62ddb452f862..768fa124105b 100644

drivers/net/ethernet/intel/ixgbe/ixgbe_common.c +0 -0
@@ -3677,10 +3677,8 @@ s32 ixgbe_host_interface_command(struct

ixgbe_hw *hw, void *buffer,
bool return_data)

 {
        u32 hdr_size = sizeof(struct ixgbe_hic_hdr);
-       union {
-               struct ixgbe_hic_hdr hdr;
-               u32 u32arr[1];
-       } *bp = buffer;
+       u32 *bp = buffer;
+       struct ixgbe_hic_hdr hdr;
        u16 buf_len, dword_len;
        s32 status;
        u32 bi;

@@ -3706,12 +3704,13 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,

        /* first pull in the header so we know the buffer length */
        for (bi = 0; bi < dword_len; bi++) {
-               bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
-               le32_to_cpus(&bp->u32arr[bi]);
+               bp[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
+               le32_to_cpus(&bp[bi]);
        }

        /* If there is any thing in data position pull it in */
-       buf_len = bp->hdr.buf_len;
+       memcpy(&hdr, bp, sizeof(hdr));
+       buf_len = hdr.buf_len;
        if (!buf_len)
                goto rel_out;

@@ -3726,8 +3725,8 @@ s32 ixgbe_host_interface_command(struct ixgbe_hw *hw, void *buffer,

        /* Pull in the rest of the buffer (bi is where we left off) */
        for (; bi <= dword_len; bi++) {
-               bp->u32arr[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
-               le32_to_cpus(&bp->u32arr[bi]);
+               bp[bi] = IXGBE_READ_REG_ARRAY(hw, IXGBE_FLEX_MNG, bi);
+               le32_to_cpus(&bp[bi]);
        }

rel_out: