[RFC PATCH 3/3] wifi: ath11k: use private page-frag caches for RXDMA rings
From: Mark Ruvald Pedersen <hidden>
Date: 2026-07-19 21:58:35
Also in:
ath11k
Subsystem:
atheros ath generic utilities, qualcomm atheros ath11k wireless driver, the rest · Maintainers:
Jeff Johnson, Linus Torvalds
ath11k allocates RXDMA skb heads with dev_alloc_skb(), which draws from
shared per-CPU page-frag caches. RX buffers from different rings and
unrelated networking allocations can therefore share a high-order
backing page despite having different ownership lifetimes.
On IPQ8074, RX ownership tracing found bounded ring and IDR occupancy
while page-frag backing grew. Two captures found current or historical
cross-origin fragments on 86-87% of the final tracker-visible physical
pages.
Give each RXDMA ring its own page_frag_cache and use it for regular and
monitor-status buffers. New fragments from distinct rings can no longer
share backing pages with one another or with unrelated users. The
monitor-destination path that intentionally uses the data refill ring
also uses that ring's cache.
Commit d455e805de70 ("wifi: ath11k: rearrange IRQ enable/disable in reset
path") moved HIF IRQ shutdown from the common crash-reconfiguration path
into the reset worker. The non-reset QMI recovery path can consequently
reach DP teardown while NAPI can still refill an RXDMA ring.
Quiesce HIF IRQ/NAPI processing on that non-reset path before teardown.
The reset worker already does so before power cycling, therefore keep the
existing reset path unchanged. This makes the cache change self-contained
without relying on repeated IRQ lifecycle transitions.
The refill paths serialize each cache with the corresponding SRNG lock.
Drain the cache only after ring-owned skbs have been DMA-unmapped and
freed and the IDR has been destroyed. Keep the dev_alloc_skb() allocation
flags and headroom behavior; build_skb() propagates head-frag and
pfmemalloc metadata from the backing page.
This leaves buffer and ring sizes and all DMA map/unmap operations
unchanged. It isolates cross-owner lifetime mixing; it does not prevent
lifetime variation among buffers belonging to the same ring.
With the private caches, page-frag backing remained bounded in natural
load, resident, and high-packet-rate pressure runs on IPQ8074. The
non-reset recovery path has not been runtime-tested in isolation.
Fixes: d5c65159f289 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Fixes: d455e805de70 ("wifi: ath11k: rearrange IRQ enable/disable in reset path")
Tested-on: IPQ8074 hw2.0 AHB WLAN.HK.2.9.0.1-02146-QCAHKSWPL_SILICONZ-1
Signed-off-by: Mark Ruvald Pedersen <redacted>
---diff --git a/drivers/net/wireless/ath/ath11k/core.c b/drivers/net/wireless/ath/ath11k/core.c
index 8dacc878c006..e71a22ce332d 100644
--- a/drivers/net/wireless/ath/ath11k/core.c
+++ b/drivers/net/wireless/ath/ath11k/core.c@@ -2330,6 +2330,9 @@ static int ath11k_core_reconfigure_on_crash(struct ath11k_base *ab) { int ret; + if (!ab->is_reset) + ath11k_hif_irq_disable(ab); + mutex_lock(&ab->core_lock); ath11k_thermal_unregister(ab); ath11k_dp_pdev_free(ab);
diff --git a/drivers/net/wireless/ath/ath11k/dp.h b/drivers/net/wireless/ath/ath11k/dp.h
index 84f66839f0c6..e69dc23b2051 100644
--- a/drivers/net/wireless/ath/ath11k/dp.h
+++ b/drivers/net/wireless/ath/ath11k/dp.h@@ -7,6 +7,8 @@ #ifndef ATH11K_DP_H #define ATH11K_DP_H +#include <linux/page_frag_cache.h> + #include "hal_rx.h" #define MAX_RXDMA_PER_PDEV 2
@@ -72,6 +74,7 @@ struct dp_srng { struct dp_rxdma_ring { struct dp_srng refill_buf_ring; + struct page_frag_cache frag_cache; struct idr bufs_idr; /* Protects bufs_idr */ spinlock_t idr_lock;
diff --git a/drivers/net/wireless/ath/ath11k/dp_rx.c b/drivers/net/wireless/ath/ath11k/dp_rx.c
index e4ea0c216740..6bbfa0ebddd0 100644
--- a/drivers/net/wireless/ath/ath11k/dp_rx.c
+++ b/drivers/net/wireless/ath/ath11k/dp_rx.c@@ -5,9 +5,11 @@ */ #include <linux/fips.h> +#include <linux/gfp.h> #include <linux/ieee80211.h> #include <linux/kernel.h> #include <linux/skbuff.h> +#include <net/sock.h> #include "core.h" #include "debug.h" #include "debugfs_htt_stats.h"
@@ -340,6 +342,36 @@ static int ath11k_dp_purge_mon_ring(struct ath11k_base *ab) return -ETIMEDOUT; } +static struct sk_buff * +ath11k_dp_rx_alloc_skb(struct dp_rxdma_ring *rx_ring, unsigned int len) +{ + struct page_frag_cache *cache = &rx_ring->frag_cache; + gfp_t gfp_mask = GFP_ATOMIC; + struct sk_buff *skb; + void *data; + + len += NET_SKB_PAD; + len = SKB_HEAD_ALIGN(len); + + if (sk_memalloc_socks()) + gfp_mask |= __GFP_MEMALLOC; + + /* Refill paths serialize the per-ring cache with the SRNG lock. */ + data = page_frag_alloc(cache, len, gfp_mask); + if (!data) + return NULL; + + skb = build_skb(data, len); + if (!skb) { + page_frag_free(data); + return NULL; + } + + skb_reserve(skb, NET_SKB_PAD); + + return skb; +} + /* Returns number of Rx buffers replenished */ int ath11k_dp_rxbufs_replenish(struct ath11k_base *ab, int mac_id, struct dp_rxdma_ring *rx_ring,
@@ -371,8 +403,8 @@ int ath11k_dp_rxbufs_replenish(struct ath11k_base *ab, int mac_id, num_remain = req_entries; while (num_remain > 0) { - skb = dev_alloc_skb(DP_RX_BUFFER_SIZE + - DP_RX_BUFFER_ALIGN_SIZE); + skb = ath11k_dp_rx_alloc_skb(rx_ring, DP_RX_BUFFER_SIZE + + DP_RX_BUFFER_ALIGN_SIZE); if (!skb) break;
@@ -453,6 +485,8 @@ static int ath11k_dp_rxdma_buf_ring_free(struct ath11k *ar, idr_destroy(&rx_ring->bufs_idr); spin_unlock_bh(&rx_ring->idr_lock); + page_frag_cache_drain(&rx_ring->frag_cache); + return 0; }
@@ -2867,8 +2901,8 @@ static struct sk_buff *ath11k_dp_rx_alloc_mon_status_buf(struct ath11k_base *ab, struct sk_buff *skb; dma_addr_t paddr; - skb = dev_alloc_skb(DP_RX_BUFFER_SIZE + - DP_RX_BUFFER_ALIGN_SIZE); + skb = ath11k_dp_rx_alloc_skb(rx_ring, DP_RX_BUFFER_SIZE + + DP_RX_BUFFER_ALIGN_SIZE); if (!skb) goto fail_alloc_skb;