From: "Eason Lai (賴易聖)" <Eason.Lai@mediatek.com>
To: "lorenzo@kernel.org" <lorenzo@kernel.org>
Cc: "Yf Luo (羅元富)" <Yf.Luo@mediatek.com>,
"nbd@nbd.name" <nbd@nbd.name>,
"Kun Wu (吳謹至)" <kun.wu@mediatek.com>,
"Quan Zhou (周全)" <Quan.Zhou@mediatek.com>,
"Deren Wu (武德仁)" <Deren.Wu@mediatek.com>,
"Ryder Lee" <Ryder.Lee@mediatek.com>,
"linux-wireless@vger.kernel.org" <linux-wireless@vger.kernel.org>,
"Sean Wang" <Sean.Wang@mediatek.com>,
"Leon Yen (顏良儒)" <Leon.Yen@mediatek.com>,
"linux-mediatek@lists.infradead.org"
<linux-mediatek@lists.infradead.org>,
"Litien Chang (張立典)" <Litien.Chang@mediatek.com>,
"JB Tsai (蔡志彬)" <Jb.Tsai@mediatek.com>
Subject: Re: [PATCH 1/3] wifi: mt76: Separate skb and page_pool_buf pointers in mt76_txwi_cache
Date: Mon, 13 Jul 2026 01:57:15 +0000 [thread overview]
Message-ID: <b2c7f6c74536295cea74dbab8ff815d60b4a747e.camel@mediatek.com> (raw)
In-Reply-To: <ak9Zjm-czxVIjrhX@lore-desk>
On Thu, 2026-07-09 at 10:19 +0200, Lorenzo Bianconi wrote:
> > From: Eason Lai <Eason.Lai@mediatek.com>
> >
> > Refactor mt76_txwi_cache structure to use separate skb and
> > page_pool_buf
> > fields instead of a union with ptr. This improves type safety and
> > makes
> > the code more explicit about which pointer type is being used in
> > different contexts.
> >
> > Also add skip_unmap flag to tx_info.buf to handle cases where DMA
> > unmapping should be skipped.
> >
> > Signed-off-by: Eason Lai <Eason.Lai@mediatek.com>
> > ---
> > drivers/net/wireless/mediatek/mt76/dma.c | 29 ++++++++++++---
> > ----
> > drivers/net/wireless/mediatek/mt76/mt76.h | 4 +--
> > .../net/wireless/mediatek/mt76/mt7996/mac.c | 10 +++----
> > drivers/net/wireless/mediatek/mt76/tx.c | 2 +-
> > drivers/net/wireless/mediatek/mt76/wed.c | 6 ++--
> > 5 files changed, 28 insertions(+), 23 deletions(-)
> >
> > diff --git a/drivers/net/wireless/mediatek/mt76/dma.c
> > b/drivers/net/wireless/mediatek/mt76/dma.c
> > index f8c2fe5f2f58..2716278788bd 100644
> > --- a/drivers/net/wireless/mediatek/mt76/dma.c
> > +++ b/drivers/net/wireless/mediatek/mt76/dma.c
> > @@ -43,7 +43,8 @@ mt76_alloc_rxwi(struct mt76_dev *dev)
> > if (!t)
> > return NULL;
> >
> > - t->ptr = NULL;
> > + t->skb = NULL;
> > + t->page_pool_buf = NULL;
> > return t;
> > }
> >
> > @@ -84,8 +85,11 @@ mt76_get_txwi(struct mt76_dev *dev)
> > {
> > struct mt76_txwi_cache *t = __mt76_get_txwi(dev);
> >
> > - if (t)
> > + if (t) {
> > + t->skb = NULL;
> > + t->page_pool_buf = NULL;
> > return t;
> > + }
> >
> > return mt76_alloc_txwi(dev);
> > }
> > @@ -147,8 +151,8 @@ mt76_free_pending_rxwi(struct mt76_dev *dev)
> >
> > local_bh_disable();
> > while ((t = __mt76_get_rxwi(dev)) != NULL) {
> > - if (t->ptr)
> > - mt76_put_page_pool_buf(t->ptr, false);
> > + if (t->page_pool_buf)
> > + mt76_put_page_pool_buf(t->page_pool_buf,
> > false);
> > kfree(t);
> > }
> > local_bh_enable();
> > @@ -475,14 +479,14 @@ mt76_dma_get_rxdmad_c_buf(struct mt76_dev
> > *dev, struct mt76_queue *q,
> > if (more)
> > *more = !FIELD_GET(RRO_RXDMAD_DATA1_LS_MASK,
> > data1);
> >
> > - buf = t->ptr;
> > + buf = t->page_pool_buf;
> > ind_reason = FIELD_GET(RRO_RXDMAD_DATA2_IND_REASON_MASK,
> > data2);
> > if (ind_reason == MT_DMA_WED_IND_REASON_REPEAT ||
> > ind_reason == MT_DMA_WED_IND_REASON_OLDPKT) {
> > mt76_put_page_pool_buf(buf, false);
> > buf = ERR_PTR(-EAGAIN);
> > }
> > - t->ptr = NULL;
> > + t->page_pool_buf = NULL;
> > t->dma_addr = 0;
> >
> > mt76_put_rxwi(dev, t);
> > @@ -529,9 +533,9 @@ mt76_dma_get_buf(struct mt76_dev *dev, struct
> > mt76_queue *q, int idx,
> > SKB_WITH_OVERHEAD(q->buf_size),
> > page_pool_get_dma_dir(q-
> > >page_pool));
> >
> > - buf = t->ptr;
> > + buf = t->page_pool_buf;
> > t->dma_addr = 0;
> > - t->ptr = NULL;
> > + t->page_pool_buf = NULL;
> >
> > mt76_put_rxwi(dev, t);
> > if (drop)
> > @@ -694,6 +698,7 @@ mt76_dma_tx_queue_skb(struct mt76_phy *phy,
> > struct mt76_queue *q,
> > goto unmap;
> >
> > tx_info.buf[n].addr = addr;
> > + tx_info.buf[n].skip_unmap = false;
> > tx_info.buf[n++].len = iter->len;
> > }
> > tx_info.nbuf = n;
> > @@ -718,9 +723,11 @@ mt76_dma_tx_queue_skb(struct mt76_phy *phy,
> > struct mt76_queue *q,
> > tx_info.info, tx_info.skb, t);
> >
> > unmap:
> > - for (n--; n > 0; n--)
> > - dma_unmap_single(dev->dma_dev,
> > tx_info.buf[n].addr,
> > - tx_info.buf[n].len,
> > DMA_TO_DEVICE);
> > + for (n--; n > 0; n--) {
> > + if (!tx_info.buf[n].skip_unmap)
> > + dma_unmap_single(dev->dma_dev,
> > tx_info.buf[n].addr,
> > + tx_info.buf[n].len,
> > DMA_TO_DEVICE);
> > + }
> >
> > free:
> > #ifdef CONFIG_NL80211_TESTMODE
> > diff --git a/drivers/net/wireless/mediatek/mt76/mt76.h
> > b/drivers/net/wireless/mediatek/mt76/mt76.h
> > index 527bef97e122..927c21536f4e 100644
> > --- a/drivers/net/wireless/mediatek/mt76/mt76.h
> > +++ b/drivers/net/wireless/mediatek/mt76/mt76.h
> > @@ -445,10 +445,8 @@ struct mt76_txwi_cache {
> > struct list_head list;
> > dma_addr_t dma_addr;
> >
> > - union {
> > struct sk_buff *skb;
> > - void *ptr;
> > - };
> > + void *page_pool_buf;
>
> if you do not rename the pointer here, the patch would be much less
> intrusive.
>
> Regards,
> Lorenzo
>
OK, I will revert the original naming in V2.
> >
> > u8 qid;
> > u8 phy_idx;
> > diff --git a/drivers/net/wireless/mediatek/mt76/mt7996/mac.c
> > b/drivers/net/wireless/mediatek/mt76/mt7996/mac.c
> > index e2a83da3a09c..924b0dc0ff1e 100644
> > --- a/drivers/net/wireless/mediatek/mt76/mt7996/mac.c
> > +++ b/drivers/net/wireless/mediatek/mt76/mt7996/mac.c
> > @@ -1775,12 +1775,12 @@ static void mt7996_rx_token_put(struct
> > mt7996_dev *dev)
> > struct mt76_txwi_cache *t;
> >
> > t = mt76_rx_token_release(&dev->mt76, i);
> > - if (!t || !t->ptr)
> > + if (!t || !t->page_pool_buf)
> > continue;
> >
> > - mt76_put_page_pool_buf(t->ptr, false);
> > + mt76_put_page_pool_buf(t->page_pool_buf, false);
> > t->dma_addr = 0;
> > - t->ptr = NULL;
> > + t->page_pool_buf = NULL;
> >
> > mt76_put_rxwi(&dev->mt76, t);
> > }
> > @@ -1928,14 +1928,14 @@ void mt7996_rro_rx_process(struct mt76_dev
> > *mdev, void *data)
> > goto next_page;
> >
> > qid = t->qid;
> > - buf = t->ptr;
> > + buf = t->page_pool_buf;
> > q = &mdev->q_rx[qid];
> > dma_sync_single_for_cpu(mdev->dma_dev, t-
> > >dma_addr,
> > SKB_WITH_OVERHEAD(
> > q->buf_size),
> > page_pool_get_dma_
> > dir(q->page_pool));
> >
> > t->dma_addr = 0;
> > - t->ptr = NULL;
> > + t->page_pool_buf = NULL;
> > mt76_put_rxwi(mdev, t);
> > if (!buf)
> > goto next_page;
> > diff --git a/drivers/net/wireless/mediatek/mt76/tx.c
> > b/drivers/net/wireless/mediatek/mt76/tx.c
> > index 22f9690634c9..665156a7ea65 100644
> > --- a/drivers/net/wireless/mediatek/mt76/tx.c
> > +++ b/drivers/net/wireless/mediatek/mt76/tx.c
> > @@ -899,7 +899,7 @@ int mt76_rx_token_consume(struct mt76_dev *dev,
> > void *ptr,
> > token = idr_alloc(&dev->rx_token, t, 0, dev-
> > >rx_token_size,
> > GFP_ATOMIC);
> > if (token >= 0) {
> > - t->ptr = ptr;
> > + t->page_pool_buf = ptr;
> > t->dma_addr = phys;
> > }
> > spin_unlock_bh(&dev->rx_token_lock);
> > diff --git a/drivers/net/wireless/mediatek/mt76/wed.c
> > b/drivers/net/wireless/mediatek/mt76/wed.c
> > index ed657d952de2..e1cf81d722b8 100644
> > --- a/drivers/net/wireless/mediatek/mt76/wed.c
> > +++ b/drivers/net/wireless/mediatek/mt76/wed.c
> > @@ -15,11 +15,11 @@ void mt76_wed_release_rx_buf(struct
> > mtk_wed_device *wed)
> > struct mt76_txwi_cache *t;
> >
> > t = mt76_rx_token_release(dev, i);
> > - if (!t || !t->ptr)
> > + if (!t || !t->page_pool_buf)
> > continue;
> >
> > - mt76_put_page_pool_buf(t->ptr, false);
> > - t->ptr = NULL;
> > + mt76_put_page_pool_buf(t->page_pool_buf, false);
> > + t->page_pool_buf = NULL;
> >
> > mt76_put_rxwi(dev, t);
> > }
> > --
> > 2.45.2
> >
next prev parent reply other threads:[~2026-07-13 1:57 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-09 7:55 [PATCH 0/3] Fix TX performance degradation when IOMMU is enabled Eason Lai
2026-07-09 7:55 ` [PATCH 1/3] wifi: mt76: Separate skb and page_pool_buf pointers in mt76_txwi_cache Eason Lai
2026-07-09 8:19 ` Lorenzo Bianconi
2026-07-13 1:57 ` Eason Lai (賴易聖) [this message]
2026-07-09 7:55 ` [PATCH 2/3] wifi: mt76: mt792x: Add TX page pool support for IOMMU-enabled systems Eason Lai
2026-07-09 8:27 ` Lorenzo Bianconi
2026-07-13 3:13 ` Eason Lai (賴易聖)
2026-07-09 7:55 ` [PATCH 3/3] wifi: mt76: mt792x: Restrict TX page pool to MT8196 platform Eason Lai
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=b2c7f6c74536295cea74dbab8ff815d60b4a747e.camel@mediatek.com \
--to=eason.lai@mediatek.com \
--cc=Deren.Wu@mediatek.com \
--cc=Jb.Tsai@mediatek.com \
--cc=Leon.Yen@mediatek.com \
--cc=Litien.Chang@mediatek.com \
--cc=Quan.Zhou@mediatek.com \
--cc=Ryder.Lee@mediatek.com \
--cc=Sean.Wang@mediatek.com \
--cc=Yf.Luo@mediatek.com \
--cc=kun.wu@mediatek.com \
--cc=linux-mediatek@lists.infradead.org \
--cc=linux-wireless@vger.kernel.org \
--cc=lorenzo@kernel.org \
--cc=nbd@nbd.name \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox