| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <aRxPOLxQkOUCDK1B@FUE-ALEWI-WINX>
Date: Tue, 18 Nov 2025 11:49:28 +0100
From: Alexander Wilhelm <alexander.wilhelm@...termo.com>
To: Baochen Qiang <baochen.qiang@....qualcomm.com>
Cc: Jeff Johnson <jjohnson@...nel.org>, linux-wireless@...r.kernel.org,
ath12k@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2] wifi: ath12k: fix endianness handling for SRNG ring
pointer accesses
On Tue, Nov 18, 2025 at 06:43:44PM +0800, Baochen Qiang wrote:
>
>
> On 11/18/2025 6:17 PM, Alexander Wilhelm wrote:
> > The SRNG head and tail ring pointers are stored in device memory as
> > little-endian values. On big-endian systems, direct dereferencing of these
> > pointers leads to incorrect values being read or written, causing ring
> > management issues and potentially breaking data flow.
> >
> > This patch ensures all accesses to SRNG ring pointers use the appropriate
> > endianness conversions. This affects both read and write paths for source
> > and destination rings, as well as debug output. The changes guarantee
> > correct operation on both little- and big-endian architectures.
> >
> > Signed-off-by: Alexander Wilhelm <alexander.wilhelm@...termo.com>
> > ---
> > Changes in v2:
> > - Set '__le32 *' type for 'hp_addr/tp_addr' in both 'dst_ring' and 'src_ring'
> > ---
> > drivers/net/wireless/ath/ath12k/hal.c | 35 +++++++++++++++------------
> > drivers/net/wireless/ath/ath12k/hal.h | 8 +++---
> > 2 files changed, 24 insertions(+), 19 deletions(-)
> >
> > diff --git a/drivers/net/wireless/ath/ath12k/hal.c b/drivers/net/wireless/ath/ath12k/hal.c
> > index 6406fcf5d69f..bd4d1de9eb1a 100644
> > --- a/drivers/net/wireless/ath/ath12k/hal.c
> > +++ b/drivers/net/wireless/ath/ath12k/hal.c
> > @@ -2007,7 +2007,7 @@ int ath12k_hal_srng_dst_num_free(struct ath12k_base *ab, struct hal_srng *srng,
> > tp = srng->u.dst_ring.tp;
> >
> > if (sync_hw_ptr) {
> > - hp = *srng->u.dst_ring.hp_addr;
> > + hp = le32_to_cpu(*srng->u.dst_ring.hp_addr);
> > srng->u.dst_ring.cached_hp = hp;
> > } else {
> > hp = srng->u.dst_ring.cached_hp;
> > @@ -2030,7 +2030,7 @@ int ath12k_hal_srng_src_num_free(struct ath12k_base *ab, struct hal_srng *srng,
> > hp = srng->u.src_ring.hp;
> >
> > if (sync_hw_ptr) {
> > - tp = *srng->u.src_ring.tp_addr;
> > + tp = le32_to_cpu(*srng->u.src_ring.tp_addr);
> > srng->u.src_ring.cached_tp = tp;
> > } else {
> > tp = srng->u.src_ring.cached_tp;
> > @@ -2149,9 +2149,9 @@ void ath12k_hal_srng_access_begin(struct ath12k_base *ab, struct hal_srng *srng)
> >
> > if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
> > srng->u.src_ring.cached_tp =
> > - *(volatile u32 *)srng->u.src_ring.tp_addr;
> > + le32_to_cpu(*(volatile u32 *)srng->u.src_ring.tp_addr);
>
> s/volatile u32 */volatile __le32 */ ?
I got it. I'll fix it in v3.
> > } else {
> > - hp = READ_ONCE(*srng->u.dst_ring.hp_addr);
> > + hp = le32_to_cpu(READ_ONCE(*srng->u.dst_ring.hp_addr));
> >
> > if (hp != srng->u.dst_ring.cached_hp) {
> > srng->u.dst_ring.cached_hp = hp;
> > @@ -2175,25 +2175,28 @@ void ath12k_hal_srng_access_end(struct ath12k_base *ab, struct hal_srng *srng)
> > * hence written to a shared memory location that is read by FW
> > */
> > if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
> > - srng->u.src_ring.last_tp =
> > - *(volatile u32 *)srng->u.src_ring.tp_addr;
> > + srng->u.src_ring.last_tp = le32_to_cpu(
> > + *(volatile u32 *)srng->u.src_ring.tp_addr);
>
> s/volatile u32 */volatile __le32 */ ?
Same as above, sure!
> > /* Make sure descriptor is written before updating the
> > * head pointer.
> > */
> > dma_wmb();
> > - WRITE_ONCE(*srng->u.src_ring.hp_addr, srng->u.src_ring.hp);
> > + WRITE_ONCE(*srng->u.src_ring.hp_addr,
> > + cpu_to_le32(srng->u.src_ring.hp));
> > } else {
> > - srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
> > + srng->u.dst_ring.last_hp =
> > + le32_to_cpu(*srng->u.dst_ring.hp_addr);
> > /* Make sure descriptor is read before updating the
> > * tail pointer.
> > */
> > dma_mb();
> > - WRITE_ONCE(*srng->u.dst_ring.tp_addr, srng->u.dst_ring.tp);
> > + WRITE_ONCE(*srng->u.dst_ring.tp_addr,
> > + cpu_to_le32(srng->u.dst_ring.tp));
> > }
> > } else {
> > if (srng->ring_dir == HAL_SRNG_DIR_SRC) {
> > - srng->u.src_ring.last_tp =
> > - *(volatile u32 *)srng->u.src_ring.tp_addr;
> > + srng->u.src_ring.last_tp = le32_to_cpu(
> > + *(volatile u32 *)srng->u.src_ring.tp_addr);
Same as above, sure!
> > /* Assume implementation use an MMIO write accessor
> > * which has the required wmb() so that the descriptor
> > * is written before the updating the head pointer.
> > @@ -2203,7 +2206,8 @@ void ath12k_hal_srng_access_end(struct ath12k_base *ab, struct hal_srng *srng)
> > (unsigned long)ab->mem,
> > srng->u.src_ring.hp);
> > } else {
> > - srng->u.dst_ring.last_hp = *srng->u.dst_ring.hp_addr;
> > + srng->u.dst_ring.last_hp =
> > + le32_to_cpu(*srng->u.dst_ring.hp_addr);
> > /* Make sure descriptor is read before updating the
> > * tail pointer.
> > */
> > @@ -2547,7 +2551,7 @@ void ath12k_hal_srng_shadow_update_hp_tp(struct ath12k_base *ab,
> > * HP only when then ring isn't' empty.
> > */
> > if (srng->ring_dir == HAL_SRNG_DIR_SRC &&
> > - *srng->u.src_ring.tp_addr != srng->u.src_ring.hp)
> > + le32_to_cpu(*srng->u.src_ring.tp_addr) != srng->u.src_ring.hp)
> > ath12k_hal_srng_access_end(ab, srng);
> > }
> >
> > @@ -2648,14 +2652,15 @@ void ath12k_hal_dump_srng_stats(struct ath12k_base *ab)
> > "src srng id %u hp %u, reap_hp %u, cur tp %u, cached tp %u last tp %u napi processed before %ums\n",
> > srng->ring_id, srng->u.src_ring.hp,
> > srng->u.src_ring.reap_hp,
> > - *srng->u.src_ring.tp_addr, srng->u.src_ring.cached_tp,
> > + __le32_to_cpu(*srng->u.src_ring.tp_addr),
> > + srng->u.src_ring.cached_tp,
> > srng->u.src_ring.last_tp,
> > jiffies_to_msecs(jiffies - srng->timestamp));
> > else if (srng->ring_dir == HAL_SRNG_DIR_DST)
> > ath12k_err(ab,
> > "dst srng id %u tp %u, cur hp %u, cached hp %u last hp %u napi processed before %ums\n",
> > srng->ring_id, srng->u.dst_ring.tp,
> > - *srng->u.dst_ring.hp_addr,
> > + __le32_to_cpu(*srng->u.dst_ring.hp_addr),
>
> still my v1 comment does not get addressed:
>
>
> why __le32_to_cpu() only in logging, while le32_to_cpu() elsewhere?
Sorry, I was confused with the previous issue. Yes, I saw this form on an
upstream patch where '__le32_to_cpu()' was used instead of 'le32_to_cpu()'.
Which one should I prefer? I'll unify that in v3.
Best regards
Alexander Wilhelm
Powered by blists - more mailing lists