Re: [PATCH net] net: pse-pd: pd692x0: Return fw_upload_err codes from firmware upload ops
From: Ian Rogers
Date: Fri Nov 20 2026 - 12:36:10 EST
On Thu, Nov 19, 2026 at 9:26 AM Kory Maincent <kory.maincent@xxxxxxxxxxx> wrote:
>
>
>
> On 11/19/26 18:10, Ian Rogers wrote:
> > pd692x0_fw_reset() and pd692x0_fw_write() return negative errno values
> > from i2c_master_send() directly as enum fw_upload_err instead of
> > FW_UPLOAD_ERR_RW_ERROR, and pd692x0_fw_recv_resp() ignores negative
> > errors from its trailing i2c_master_recv(). When a negative errno is
> > propagated to the firmware upload core, fw_upload_error() rejects it
> > as >= FW_UPLOAD_ERR_MAX, logs an invalid error code message, and
> > reports "unknown-error" instead of "read-write-error" via sysfs.
> >
> > Return FW_UPLOAD_ERR_RW_ERROR on I2C transfer failures and use
> > enum fw_upload_err consistently across the PD692x0 firmware upload
> > helpers.
>
> Please write a readable commit message with human hands.
>
> > Found via Clang -Wimplicit-int-enum-cast on an x86_64 allyesconfig
> > build. Verified by building drivers/net/pse-pd/pd692x0.o with LLVM=1
> > W=1 (-Werror) and -Wimplicit-int-enum-cast, and with a standalone C
> > reproducer; not runtime-tested on PD692x0 hardware.
> >
> > Fixes: 9a9938451890 ("net: pse-pd: Add PD692x0 PSE controller driver")
> > Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
> > ---
> > drivers/net/pse-pd/pd692x0.c | 81 ++++++++++++++++++------------------
> > 1 file changed, 40 insertions(+), 41 deletions(-)
> >
> > diff --git a/drivers/net/pse-pd/pd692x0.c b/drivers/net/pse-pd/pd692x0.c
> > index 7958f5741463..94a857eaa3de 100644
> > --- a/drivers/net/pse-pd/pd692x0.c
> > +++ b/drivers/net/pse-pd/pd692x0.c
> > @@ -1455,6 +1455,8 @@ pd692x0_fw_recv_resp(const struct i2c_client *client, unsigned long ms_timeout,
> >
> > /* Read remaining characters */
> > ret = i2c_master_recv(client, fw_msg_buf + 1, msg_size - 1);
> > + if (ret < 0)
> > + return FW_UPLOAD_ERR_RW_ERROR;
> > if (strncmp(fw_msg_buf, msg_ok, msg_size)) {
>
> The error will be catch anyway by the string comparison. If you want to
> add a check on ret then add another error message.
> This is not a fix.
>
> > dev_err(&client->dev,
> > "Wrong FW download process answer (%*pE)\n",
> > @@ -1465,9 +1467,10 @@ pd692x0_fw_recv_resp(const struct i2c_client *client, unsigned long ms_timeout,
> > return FW_UPLOAD_ERR_NONE;
> > }
> >
> > -static int pd692x0_fw_write_line(const struct i2c_client *client,
> > - const char line[PD692X0_FW_LINE_MAX_SZ],
> > - const bool last_line)
> > +static enum fw_upload_err
> > +pd692x0_fw_write_line(const struct i2c_client *client,
> > + const char line[PD692X0_FW_LINE_MAX_SZ],
> > + const bool last_line)
> > {
> > int ret;
> >
> > @@ -1478,19 +1481,12 @@ static int pd692x0_fw_write_line(const struct i2c_client *client,
> > line++;
> > }
> >
> > - if (last_line) {
> > - ret = pd692x0_fw_recv_resp(client, 100, "TP\r\n",
> > - sizeof("TP\r\n") - 1);
> > - if (ret)
> > - return ret;
> > - } else {
> > - ret = pd692x0_fw_recv_resp(client, 100, "T*\r\n",
> > - sizeof("T*\r\n") - 1);
> > - if (ret)
> > - return ret;
> > - }
> > + if (last_line)
> > + return pd692x0_fw_recv_resp(client, 100, "TP\r\n",
> > + sizeof("TP\r\n") - 1);
> >
> > - return FW_UPLOAD_ERR_NONE;
> > + return pd692x0_fw_recv_resp(client, 100, "T*\r\n",
> > + sizeof("T*\r\n") - 1);
>
> This is not a fix, just moving code around.
>
> For the rest, FW_UPLOAD_ERR_NONE is similar to ret = 0 so there is no
> real issue, only some type mismatch and I am not sure these changes
> should land as a fix.
Sorry for the noisy diff and dense commit message - you are right that
pd692x0_fw_recv_resp() (where fw_msg_buf is zero-initialized) and
pd692x0_fw_write_line() didn't need touching, and all that type churn
buried the actual change.
The only two lines that return a negative errno (-EIO) as an enum
fw_upload_err instead of FW_UPLOAD_ERR_RW_ERROR are:
1. pd692x0_fw_reset(): 'return ret;' when sending "RST" fails
2. pd692x0_fw_write(): 'return ret;' when sending 'P' fails (whereas
sending 'E' right above it already returns FW_UPLOAD_ERR_RW_ERROR)
Would you welcome a minimal 2-line patch for net-next (without Fixes:)
just changing those two returns to FW_UPLOAD_ERR_RW_ERROR, or would
you prefer leaving it as-is?
Thanks,
Ian
> Regards
> --
> Köry Maincent, Bootlin
> Embedded Linux and kernel engineering
> https://bootlin.com
>