On Wed, Aug 9, 2023 at 10:40 AM Xu Yang <xu.yang_2@xxxxxxx> wrote:
>
> Some NXP processor using chipidea IP has a bug when frame babble is
> detected.
>
> As per 4.15.1.1.1 Serial Bus Babble:
> A babble condition also exists if IN transaction is in progress at
> High-speed SOF2 point. This is called frame babble. The host controller
> must disable the port to which the frame babble is detected.
>
> The USB controller has disabled the port (PE cleared) and has asserted
> USBERRINT when frame babble is detected, but PEC is not asserted.
> Therefore, the SW isn't aware that port has been disabled. Then the
> SW keeps sending packets to this port, but all of the transfers will
> fail.
>
> This workaround will firstly assert PCD by SW when USBERRINT is detected
> and then judge whether port change has really occurred or not by polling
> roothub status. Because the PEC doesn't get asserted in our case, this
> patch will also assert it by SW when specific conditions are satisfied.
>
> Signed-off-by: Xu Yang <xu.yang_2@xxxxxxx>
>
Acked-by: Peter Chen <peter.chen@xxxxxxxxxx>
Peter
> ---
> Changes in v2:
> - fix typo in commit message
> - Adjust the judgement logic in echi_irq() as suggested from Alan
> ---
> drivers/usb/host/ehci-hcd.c | 8 ++++++--
> drivers/usb/host/ehci-hub.c | 10 +++++++++-
> drivers/usb/host/ehci.h | 10 ++++++++++
> 3 files changed, 25 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/usb/host/ehci-hcd.c b/drivers/usb/host/ehci-hcd.c
> index a1930db0da1c..802bfafb1012 100644
> --- a/drivers/usb/host/ehci-hcd.c
> +++ b/drivers/usb/host/ehci-hcd.c
> @@ -755,10 +755,14 @@ static irqreturn_t ehci_irq (struct usb_hcd *hcd)
>
> /* normal [4.15.1.2] or error [4.15.1.1] completion */
> if (likely ((status & (STS_INT|STS_ERR)) != 0)) {
> - if (likely ((status & STS_ERR) == 0))
> + if (likely ((status & STS_ERR) == 0)) {
> INCR(ehci->stats.normal);
> - else
> + } else {
> + /* Force to check port status */
> + if (ehci->has_ci_pec_bug)
> + status |= STS_PCD;
> INCR(ehci->stats.error);
> + }
> bh = 1;
> }
>
> diff --git a/drivers/usb/host/ehci-hub.c b/drivers/usb/host/ehci-hub.c
> index efe30e3be22f..1aee392e8492 100644
> --- a/drivers/usb/host/ehci-hub.c
> +++ b/drivers/usb/host/ehci-hub.c
> @@ -674,7 +674,8 @@ ehci_hub_status_data (struct usb_hcd *hcd, char *buf)
>
> if ((temp & mask) != 0 || test_bit(i, &ehci->port_c_suspend)
> || (ehci->reset_done[i] && time_after_eq(
> - jiffies, ehci->reset_done[i]))) {
> + jiffies, ehci->reset_done[i]))
> + || ehci_has_ci_pec_bug(ehci, temp)) {
> if (i < 7)
> buf [0] |= 1 << (i + 1);
> else
> @@ -875,6 +876,13 @@ int ehci_hub_control(
> if (temp & PORT_PEC)
> status |= USB_PORT_STAT_C_ENABLE << 16;
>
> + if (ehci_has_ci_pec_bug(ehci, temp)) {
> + status |= USB_PORT_STAT_C_ENABLE << 16;
> + ehci_info(ehci,
> + "PE is cleared by HW port:%d PORTSC:%08x\n",
> + wIndex + 1, temp);
> + }
> +
> if ((temp & PORT_OCC) && (!ignore_oc && !ehci->spurious_oc)){
> status |= USB_PORT_STAT_C_OVERCURRENT << 16;
>
> diff --git a/drivers/usb/host/ehci.h b/drivers/usb/host/ehci.h
> index c5c7f8782549..1441e3400796 100644
> --- a/drivers/usb/host/ehci.h
> +++ b/drivers/usb/host/ehci.h
> @@ -207,6 +207,7 @@ struct ehci_hcd { /* one per controller */
> unsigned has_fsl_port_bug:1; /* FreeScale */
> unsigned has_fsl_hs_errata:1; /* Freescale HS quirk */
> unsigned has_fsl_susp_errata:1; /* NXP SUSP quirk */
> + unsigned has_ci_pec_bug:1; /* ChipIdea PEC bug */
> unsigned big_endian_mmio:1;
> unsigned big_endian_desc:1;
> unsigned big_endian_capbase:1;
> @@ -707,6 +708,15 @@ ehci_port_speed(struct ehci_hcd *ehci, unsigned int portsc)
> */
> #define ehci_has_fsl_susp_errata(e) ((e)->has_fsl_susp_errata)
>
> +/*
> + * Some Freescale/NXP processors using ChipIdea IP have a bug in which
> + * disabling the port (PE is cleared) does not cause PEC to be asserted
> + * when frame babble is detected.
> + */
> +#define ehci_has_ci_pec_bug(e, portsc) \
> + ((e)->has_ci_pec_bug && ((e)->command & CMD_PSE) \
> + && !(portsc & PORT_PEC) && !(portsc & PORT_PE))
> +
> /*
> * While most USB host controllers implement their registers in
> * little-endian format, a minority (celleb companion chip) implement
> --
> 2.34.1
>
