Skip to contentSkip to conversation

wifi: brcmfmac: Fix allocation size

PATCH In review Alexey V. Vissarionov 4 participants 7 messages Jan 17, 2023 – Jan 18, 2023

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 wifi: brcmfmac: Fix allocation size

The "pkt" is a pointer to struct sk_buff, so it's just 4 or 8 bytes, while the structure itself is much bigger.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes bbd1f932e7c45ef1 ("brcmfmac: cleanup ampdu-rx host reorder code")
drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c +1 -1
index 36af81975855c525..0d283456da331464 100644
@@ -1711,7 +1711,7 @@ void brcmf_fws_rxreorder(struct brcmf_if *ifp, struct sk_buff *pkt)
 		buf_size = sizeof(*rfi);
 		max_idx = reorder_data[BRCMF_RXREORDER_MAXIDX_OFFSET];
 
-		buf_size += (max_idx + 1) * sizeof(pkt);
+		buf_size += (max_idx + 1) * sizeof(struct sk_buff);
 
 		/* allocate space for flow reorder info */
 		brcmf_dbg(INFO, "flow-%d: start, maxidx %d\n",
Signature · 3 lines
Alexey V. Vissarionov
gremlin ПРИ altlinux ТЧК org; +vii-cmiii-ccxxix-lxxix-xlii
GPG: 0D92F19E1C0DC36E27F61A29CD17E2B43D879005 @ hkp://keys.gnupg.net
#1 lore raw
Quoted 15 lines · Alexey V. Vissarionov

The "pkt" is a pointer to struct sk_buff, so it's just 4 or 8 bytes, while the structure itself is much bigger.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes bbd1f932e7c45ef1 ("brcmfmac: cleanup ampdu-rx host reorder code")

diff --git a/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c index 36af81975855c525..0d283456da331464 100644

drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c void brcmf_fws_rxreorder(struct brcmf_if *ifp, · line 1715

Quoted 6 lines
drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c +1 -1
@@ -1711,7 +1711,7 @@ void brcmf_fws_rxreorder(struct brcmf_if *ifp,
 struct sk_buff *pkt)
 		buf_size = sizeof(*rfi);
 		max_idx = reorder_data[BRCMF_RXREORDER_MAXIDX_OFFSET];
 
-		buf_size += (max_idx + 1) * sizeof(pkt);
+		buf_size += (max_idx + 1) * sizeof(struct sk_buff);

Wouldn't sizeof(*pkt) be better? Just like with sizeof(*rfi) few lines above.

Signature · 3 lines
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
#2 lore raw
Quoted 3 lines · Kalle Valo
Quoted 2 lines

- buf_size += (max_idx + 1) * sizeof(pkt);
+ buf_size += (max_idx + 1) * sizeof(struct sk_buff);

Wouldn't sizeof(*pkt) be better?

Usually sizeof(type) produces less errors than sizeof(var)...

Quoted 1 line

Just like with sizeof(*rfi) few lines above.

... but to keep consistency sizeof(*pkt) would also be ok.

Signature · 3 lines
Alexey V. Vissarionov
gremlin ПРИ altlinux ТЧК org; +vii-cmiii-ccxxix-lxxix-xlii
GPG: 0D92F19E1C0DC36E27F61A29CD17E2B43D879005 @ hkp://keys.gnupg.net
#3 lore raw
Quoted 7 lines · Alexey V. Vissarionov
Quoted 3 lines · Kalle Valo
Quoted 2 lines

- buf_size += (max_idx + 1) * sizeof(pkt);
+ buf_size += (max_idx + 1) * sizeof(struct sk_buff);

Wouldn't sizeof(*pkt) be better?

Usually sizeof(type) produces less errors than sizeof(var)...

This matter of taste really but FWIW I prefer sizeof(var) as then the type can't be different by accident. And the coding style says something similar, although that's related to memory allocation:

https://www.kernel.org/doc/html/latest/process/coding-style.html#allocating-memory

Signature · 3 lines
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
#4 lore raw
In reply to Alexey V. Vissarionov
Quoted 13 lines · Alexey V. Vissarionov

The "pkt" is a pointer to struct sk_buff, so it's just 4 or 8 bytes, while the structure itself is much bigger.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes bbd1f932e7c45ef1 ("brcmfmac: cleanup ampdu-rx host reorder code")
drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c +0 -0
index 36af81975855c525..0d283456da331464 100644

drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c void brcmf_fws_rxreorder(struct brcmf_if *ifp, struct sk_buff *pkt) · line 1717

Quoted 8 lines
drivers/net/wireless/broadcom/brcm80211/brcmfmac/fwsignal.c +1 -1
@@ -1711,7 +1711,7 @@ void brcmf_fws_rxreorder(struct brcmf_if *ifp, struct sk_buff *pkt)
 		buf_size = sizeof(*rfi);
 		max_idx = reorder_data[BRCMF_RXREORDER_MAXIDX_OFFSET];
 
-		buf_size += (max_idx + 1) * sizeof(pkt);
+		buf_size += (max_idx + 1) * sizeof(struct sk_buff);
 
 		/* allocate space for flow reorder info */
 		brcmf_dbg(INFO, "flow-%d: start, maxidx %d\n",

Hi Alexey,

This is followed by:

		rfi = kzalloc(buf_size, GFP_ATOMIC);
		...
		rfi->pktslots = (struct sk_buff **)(rfi + 1);

The type of rfi is struct brcmf_ampdu_rx_reorder, which looks like this:

struct brcmf_ampdu_rx_reorder {
        struct sk_buff **pktslots;                                       
	...
};

And it looks to me that pkt is used as an array of (struct sk_buff *).

So in all, it seems to me that the current code is correct.

Is there a particular code that leads you to think otherwise?

Kind regards,
Simon

#5 lore raw
Quoted 5 lines · Simon Horman
Quoted 4 lines

buf_size = sizeof(*rfi);
max_idx = reorder_data[BRCMF_RXREORDER_MAXIDX_OFFSET];
- buf_size += (max_idx + 1) * sizeof(pkt);
+ buf_size += (max_idx + 1) * sizeof(struct sk_buff);

Quoted 11 lines

This is followed by:
rfi = kzalloc(buf_size, GFP_ATOMIC);
...
rfi->pktslots = (struct sk_buff **)(rfi + 1);
The type of rfi is struct brcmf_ampdu_rx_reorder, which looks like this:
struct brcmf_ampdu_rx_reorder
{ struct sk_buff **pktslots; ... };
And it looks to me that pkt is used as an array of (struct sk_buff *).
So in all, it seems to me that the current code is correct.

So, the buf_size is a sum of sizeof(struct brcmf_ampdu_rx_reorder) and size of array of pointers... and yes, this array is filled with
pointers: rfi->pktslots[rfi->cur_idx] = pkt;

Hmmm... looks correct. Sorry for bothering.

Signature · 3 lines
Alexey V. Vissarionov
gremlin ПРИ altlinux ТЧК org; +vii-cmiii-ccxxix-lxxix-xlii
GPG: 0D92F19E1C0DC36E27F61A29CD17E2B43D879005 @ hkp://keys.gnupg.net
maintainer
#6 lore raw
Quoted 19 lines · Alexey V. Vissarionov
Quoted 5 lines · Simon Horman
Quoted 4 lines

buf_size = sizeof(*rfi);
max_idx = reorder_data[BRCMF_RXREORDER_MAXIDX_OFFSET];
- buf_size += (max_idx + 1) * sizeof(pkt);
+ buf_size += (max_idx + 1) * sizeof(struct sk_buff);

Quoted 11 lines

This is followed by:
rfi = kzalloc(buf_size, GFP_ATOMIC);
...
rfi->pktslots = (struct sk_buff **)(rfi + 1);
The type of rfi is struct brcmf_ampdu_rx_reorder, which looks like this:
struct brcmf_ampdu_rx_reorder
{ struct sk_buff **pktslots; ... };
And it looks to me that pkt is used as an array of (struct sk_buff *).
So in all, it seems to me that the current code is correct.

Quoted 5 lines

So, the buf_size is a sum of sizeof(struct brcmf_ampdu_rx_reorder) and size of array of pointers... and yes, this array is filled with
pointers: rfi->pktslots[rfi->cur_idx] = pkt;

Hmmm... looks correct. Sorry for bothering.

No problem. Nice to see the water went still without me chiming in. It has been a while since this was added to the driver and there could be issues with this code, but if this allocation was wrong we would have had reports by now.

Thanks,
Arend