From: Yongzhao Chen <yongzhao.derek@gmail.com>
To: netdev@vger.kernel.org
Cc: Christian Marangi <ansuelsmth@gmail.com>,
Andrew Lunn <andrew@lunn.ch>, Vladimir Oltean <olteanv@gmail.com>,
"David S. Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@kernel.org>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
John Crispin <john@phrozen.org>,
linux-kernel@vger.kernel.org
Subject: [PATCH net-next v3 2/3] net: dsa: qca8k: do not clear MASTER_EN after a failed page select
Date: Sat, 3 Oct 2026 19:24:03 +0200 [thread overview]
Message-ID: <20261003172405.342-3-yongzhao.derek@gmail.com> (raw)
In-Reply-To: <20261003172405.342-1-yongzhao.derek@gmail.com>
qca8k_mdio_write() and qca8k_mdio_read() take the common exit path when
selecting the page of QCA8K_MDIO_MASTER_CTRL fails. That path clears
MASTER_CTRL[31:16] with a raw write to phy 0x10, reg 0x1f, which only
addresses MASTER_CTRL while page 0 is selected. After a failed page
select that did not reach the switch, the previous page remains selected.
The write can then clear the upper half of an unrelated register at
page * 0x200 + 0x3c.
No MDIO master transaction has been started at that point, so there is
nothing to clean up. Unlock and return the error directly.
qca8k_set_page() also keeps the old cached page when the page write
fails. If the write did reach the switch, a later access to the cached
page skips the page select and reaches the wrong register. Invalidate
the cache on failure so that the next access selects the page again.
The Sashiko review of v1 identified the stray cleanup write. A userspace
model of the driver's functions reproduces both problems by injecting
page-select failures with and without a hardware page change; these cases
pass after this change.
The cleanup write comes from commit 759bafb8a322 ("net: dsa: qca8k: add
support for internal phy and internal mdio"), and the stale page cache
from commit ba5707ec58cf ("net: dsa: qca8k: handle qca8k_set_page
errors").
Signed-off-by: Yongzhao Chen <yongzhao.derek@gmail.com>
Assisted-by: LLM
---
v3: Target net-next and drop the Fixes: tags. The commits that
introduced the problems are named in the text.
v2: new patch, addressing the Sashiko review of v1:
https://lore.kernel.org/netdev/179054715863.3145.10179961093285192493@kernel.org/
drivers/net/dsa/qca/qca8k-8xxx.c | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/drivers/net/dsa/qca/qca8k-8xxx.c b/drivers/net/dsa/qca/qca8k-8xxx.c
index b73c7c52e0f..2708430413d 100644
--- a/drivers/net/dsa/qca/qca8k-8xxx.c
+++ b/drivers/net/dsa/qca/qca8k-8xxx.c
@@ -153,6 +153,8 @@ qca8k_set_page(struct qca8k_priv *priv, u16 page)
ret = bus->write(bus, 0x18, 0, page);
if (ret < 0) {
+ /* The switch may or may not have switched pages. */
+ *cached_page = 0xffff;
dev_err_ratelimited(&bus->dev,
"failed to set qca8k page\n");
return ret;
@@ -836,7 +838,7 @@ qca8k_mdio_write(struct qca8k_priv *priv, int phy, int regnum, u16 data)
ret = qca8k_set_page(priv, page);
if (ret)
- goto exit;
+ goto unlock;
ret = qca8k_mii_write32(bus, 0x10 | r2, r1, val);
if (ret < 0)
@@ -851,6 +853,7 @@ qca8k_mdio_write(struct qca8k_priv *priv, int phy, int regnum, u16 data)
if (!ret)
ret = ret1;
+unlock:
mutex_unlock(&bus->mdio_lock);
return ret;
@@ -877,7 +880,7 @@ qca8k_mdio_read(struct qca8k_priv *priv, int phy, int regnum)
ret = qca8k_set_page(priv, page);
if (ret)
- goto exit;
+ goto unlock;
ret = qca8k_mii_write_hi(bus, 0x10 | r2, r1 + 1, val);
if (ret < 0)
@@ -896,6 +899,7 @@ qca8k_mdio_read(struct qca8k_priv *priv, int phy, int regnum)
if (!ret)
ret = ret1;
+unlock:
mutex_unlock(&bus->mdio_lock);
if (ret >= 0)
--
2.43.0
next prev parent reply other threads:[~2026-10-03 17:24 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-10-03 17:24 [PATCH net-next v3 0/3] net: dsa: qca8k: fix MDIO error handling Yongzhao Chen
2026-10-03 17:24 ` [PATCH net-next v3 1/3] net: dsa: qca8k: propagate MDIO errors Yongzhao Chen
2026-10-03 17:24 ` Yongzhao Chen [this message]
2026-10-03 17:24 ` [PATCH net-next v3 3/3] net: dsa: qca8k: fail mgmt Ethernet MDIO access on busy wait errors Yongzhao Chen
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=20261003172405.342-3-yongzhao.derek@gmail.com \
--to=yongzhao.derek@gmail.com \
--cc=andrew@lunn.ch \
--cc=ansuelsmth@gmail.com \
--cc=davem@davemloft.net \
--cc=edumazet@kernel.org \
--cc=john@phrozen.org \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=olteanv@gmail.com \
--cc=pabeni@redhat.com \
/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
all inboxes | Powered by JetHome®