Re: [PATCH] net: stmmac: program the FSM timeout timer
From: zhu xin
Date: Sun Nov 15 2026 - 22:43:05 EST
Hi,
Thanks for the review. Responses below.
On Sun, Nov 15, 2026 at 10:52 PM <netdev-bot+sashiko@xxxxxxxxxx> wrote:
>
> Thank you for your contribution! Sashiko AI review found 3 potential
> issue(s) to consider.
>
> Critical: 0 · High: 2 · Medium: 1 · Low: 0
>
> If you believe none of the issues are legitimate, please reply to each
> issue of Medium severity or higher and explain why the AI is wrong.
> If any of them are legitimate and you plan to respin, please let patchwork
> know by including "pw-bot: cr" as a separate line at the end of your reply
> (one such reply per series is enough).
>
> - [High] On every in-tree XGMAC platform, the XGMAC half of this fix
> never runs.
> - [High] The new code takes the CSR clock rate only from
> priv->plat->stmmac_clk (stmmac_safety_feat_configuration(),…
> - [Medium] dwmac5_safety_feat_config() (dwmac5.c:261-263) and
> dwxgmac3_safety_feat_config() (dwxgmac2_core.c, around lines 887-889)…
>
> --
>
> net: stmmac: program the FSM timeout timer
>
> This commit programs MAC_FSM_ACT_Timer right after TMOUTEN is enabled, in
> both dwmac5_safety_feat_config() and dwxgmac3_safety_feat_config(). TMR is
> derived from the CSR clock rate and NTMRMD defaults to a 1s timeout. A new
> ntmrmd field in stmmac_safety_feature_cfg lets a platform override that
> default, and csr_clk_mhz is passed through the safety_feat_config callback.
>
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac5.c b/drivers/net/ethernet/stmicro/stmmac/dwmac5.c
> > index f1bb981cab7cd..4ab7a3df98c6d 100644
> > --- a/drivers/net/ethernet/stmicro/stmmac/dwmac5.c
> > +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac5.c
>
> [ ... ]
>
> > @@ -245,6 +247,22 @@ int dwmac5_safety_feat_config(void __iomem *ioaddr, unsigned int asp,
> > value |= TMOUTEN; /* FSM Timeout Feature */
> > writel(value, ioaddr + MAC_FSM_CONTROL);
> >
> > + /* Program the FSM active timer so the timeout checker actually fires.
> > + * TMR is the number of CSR clocks per 1us tic, i.e. csr_clk in MHz.
> > + * Default to a 1s timeout when the platform did not select one.
> > + */
> > + if (safety_feat_cfg->tmouten && csr_clk_mhz) {
> > + u32 ntmrmd = safety_feat_cfg->ntmrmd;
> > +
> > + ntmrmd = (ntmrmd >= STMMAC_NTMRMD_1US &&
> > + ntmrmd <= STMMAC_NTMRMD_64SEC) ?
> > + ntmrmd : STMMAC_NTMRMD_1SEC;
> > +
> > + writel(FIELD_PREP(MAC_FSM_ACT_TIMER_NTMRMD, ntmrmd) |
> > + FIELD_PREP(MAC_FSM_ACT_TIMER_TMR, csr_clk_mhz),
> > + ioaddr + MAC_FSM_ACT_TIMER);
>
> [Severity: Medium]
> This writel() replaces the whole MAC_FSM_ACT_Timer register with only the
> NTMRMD and TMR fields set. Every other bit becomes zero. Is that intended?
>
> In the Synopsys EQoS/XGMAC layout, this register also has a large-mode
> timeout field (LTMRMD, bits 23:20) for FSMs that need longer timeouts. The
> patch doesn't define or program it.
>
> The commit message says a zero mode field disables the timer. If that also
> applies to LTMRMD, would the large-mode FSMs stay unmonitored after this
> write? And if reset or boot firmware had set LTMRMD, would this write turn
> that monitoring off?
>
> Should this be a read-modify-write, or should LTMRMD be programmed too? The
> same whole-register write is in dwxgmac3_safety_feat_config().
>
> > + }
> > +
>
Agreed, I'll make the MAC_FSM_ACT_Timer write a read-modify-write.
> [ ... ]
>
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c b/drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c
> > index 1a88cbaed70ca..76c850ea13a5c 100644
> > --- a/drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c
> > +++ b/drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c
>
> [ ... ]
>
> > @@ -871,6 +873,22 @@ dwxgmac3_safety_feat_config(void __iomem *ioaddr, unsigned int asp,
> > value |= XGMAC_TMOUTEN; /* FSM Timeout Feature */
> > writel(value, ioaddr + XGMAC_MAC_FSM_CONTROL);
> >
> > + /* Program the FSM active timer so the timeout checker actually fires.
> > + * TMR is the number of CSR clocks per 1us tic, i.e. csr_clk in MHz.
> > + * Default to a 1s timeout when the platform did not select one.
> > + */
> > + if (safety_cfg && safety_cfg->tmouten && csr_clk_mhz) {
>
> [Severity: High]
> Can this block run on any in-tree XGMAC platform?
>
> When safety_cfg is NULL, dwmac5_safety_feat_config() substitutes a default
> config, but dwxgmac3_safety_feat_config() does not:
>
> dwmac5_safety_feat_config() {
> ...
> if (!safety_feat_cfg)
> safety_feat_cfg = &all_safety_feats;
> ...
> }
>
> stmmac_safety_feat_configuration() passes priv->plat->safety_feat_cfg
> through unchanged. The only in-tree platforms that allocate it appear to be
> dwmac-intel.c, dwmac-mediatek.c and stmmac_pci.c, and all of them are
> GMAC/GMAC4 cores.
>
> The XGMAC users leave it NULL: dwmac-socfpga.c, dwmac-tegra.c and generic
> "snps,dwxgmac" DT through stmmac_platform.c. Both dwxgmac210_ops and
> dwxlgmac2_ops use this function.
>
> On those platforms, XGMAC_TMOUTEN is still set unconditionally above, but
> the timer write is skipped. The hardware stays in the state the commit
> message describes as the bug, even though the patch carries:
>
> Fixes: 56e58d6c8a56 ("net: stmmac: Implement Safety Features in XGMAC core")
>
> This also contradicts the new comment "Default to a 1s timeout when the
> platform did not select one".
>
> There is a related mismatch when a cfg is passed with tmouten = 0: TMOUTEN
> is still set, but the timer is not programmed.
>
> Should TMOUTEN and the timer write be gated on the same condition? And
> should a NULL cfg be handled the same way dwmac5 handles it?
>
This is due to dwxgmac3_safety_feat_config() not having the
all_safety_feats fallback that dwmac5_safety_feat_config() has -- a
pre-existing difference, not introduced by this patch. Fixing it
(adding the fallback and gating XGMAC_PRTYEN/XGMAC_TMOUTEN on the
config, matching dwmac5) changes the existing unconditional behavior.
Perhaps it would be more appropriate to address this part as a separate
optimization later on.
> [ ... ]
>
> > diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> > index f8dceafdda2d2..52f63f624d67c 100644
> > --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> > +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> > @@ -3639,10 +3639,16 @@ static void stmmac_mtl_configuration(struct stmmac_priv *priv)
> >
> > static void stmmac_safety_feat_configuration(struct stmmac_priv *priv)
> > {
> > + unsigned int csr_clk_mhz = 0;
> > +
> > if (priv->dma_cap.asp) {
> > netdev_info(priv->dev, "Enabling Safety Features\n");
> > + if (priv->plat->stmmac_clk)
> > + csr_clk_mhz = clk_get_rate(priv->plat->stmmac_clk) /
> > + 1000000;
>
> [Severity: High]
> What happens when priv->plat->stmmac_clk is NULL here? csr_clk_mhz stays 0.
> By the time dwmac5_safety_feat_config() reaches the timer block it has
> already set TMOUTEN, and it then skips the MAC_FSM_ACT_Timer write without
> any message:
>
> if (safety_feat_cfg->tmouten && csr_clk_mhz) {
>
> Both in-tree platforms that explicitly set safety_feat_cfg->tmouten = 1
> appear to have no stmmac_clk.
>
> In stmmac_pci.c, stmmac_pci_probe() sets tmouten = 1 for snps_gmac5. The
> only CSR clock information is plat->clk_csr = STMMAC_CSR_250_300M in
> snps_gmac5_default_data(), and nothing in stmmac_pci.c assigns stmmac_clk.
>
> In dwmac-mediatek.c, mediatek_dwmac_common_data() sets tmouten = 1. The
> mediatek binding and dtsi files have no "stmmaceth" clock-name. As a result,
> devm_clk_get() fails in stmmac_probe_config_dt(), which prints "Cannot get
> CSR clock" and sets plat->stmmac_clk = NULL.
>
> On both platforms the timer is never programmed and nothing is logged. That
> doesn't match the new comment on stmmac_safety_feature_cfg.ntmrmd, which
> says TMR "is derived from the CSR clock by stmmac" and that "0 selects the
> default 1s timeout".
>
> Could this fall back to plat->clk_csr, use a different clock source, or at
> least warn when the timer can't be programmed?
>
> > stmmac_safety_feat_config(priv, priv->ioaddr, priv->dma_cap.asp,
> > - priv->plat->safety_feat_cfg);
> > + priv->plat->safety_feat_cfg,
> > + csr_clk_mhz);
>
Agreed, I'll add a netdev_warn() when the CSR clock is unavailable so
the skipped timer programming is not silent. I deliberately did not
fall back to plat->clk_csr: clk_csr selects the MDC divider (a
frequency band, e.g. 250-300M), it does not hold the exact CSR clock
rate, so it cannot derive TMR accurately.
> [ ... ]
>
> --
> Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261111143322.2234321-1-zxyan0222%40gmail.com
pw-bot: cr