From: Mika Westerberg <mika.westerberg@linux.intel.com>
To: Basavaraj Natikar <Basavaraj.Natikar@amd.com>
Cc: andreas.noever@gmail.com, westeri@kernel.org,
YehezkelShB@gmail.com, linux-usb@vger.kernel.org,
Mario.Limonciello@amd.com, Sanath S <Sanath.S@amd.com>
Subject: Re: [PATCH -next v2] thunderbolt: Add quirk to reset host interface on DMA path teardown for AMD USB4 routers
Date: Wed, 5 Aug 2026 14:28:32 +0200 [thread overview]
Message-ID: <20260805122832.GH235112@black.igk.intel.com> (raw)
In-Reply-To: <20260805111811.517996-1-Basavaraj.Natikar@amd.com>
Hi,
On Wed, Aug 05, 2026 at 04:48:11PM +0530, Basavaraj Natikar wrote:
> Some AMD USB4 host routers have a bug in the Host Interface where
> DMA path setup and teardown cycles may cause the Tx ring to hang.
>
> Fix this by issuing a Host Interface Reset on every DMA path teardown
> for affected routers. The Host Interface Reset brings the registers in
> the memory BAR to their default state and clears the End-to-End Flow
> Control state, preventing the hang condition.
>
> Co-developed-by: Sanath S <Sanath.S@amd.com>
> Signed-off-by: Sanath S <Sanath.S@amd.com>
> Signed-off-by: Basavaraj Natikar <Basavaraj.Natikar@amd.com>
> ---
> v2:
> - Move the quirk to the NHI (nhi->quirks), set in nhi_pci_check_quirks()
> by PCI ID following the QUIRK_AUTO_CLEAR_INT convention, instead of the
> tb_quirks[] fabric table.
> - Declare nhi_host_interface_reset() in nhi.h instead of tb.h.
>
> v1: https://lore.kernel.org/all/20260804122638.1623429-1-Basavaraj.Natikar@amd.com/
> drivers/thunderbolt/nhi.c | 37 ++++++++++++++++++++++++++++++++++
> drivers/thunderbolt/nhi.h | 2 ++
> drivers/thunderbolt/nhi_regs.h | 4 ++++
> drivers/thunderbolt/pci.c | 21 +++++++++++++++++++
> drivers/thunderbolt/tb.c | 6 ++++++
> 5 files changed, 70 insertions(+)
>
> diff --git a/drivers/thunderbolt/nhi.c b/drivers/thunderbolt/nhi.c
> index 383a36212f70..72a76db5ea68 100644
> --- a/drivers/thunderbolt/nhi.c
> +++ b/drivers/thunderbolt/nhi.c
> @@ -1160,6 +1160,43 @@ static void nhi_reset(struct tb_nhi *nhi)
> dev_warn(nhi->dev, "timeout resetting host router\n");
> }
>
> +/**
> + * nhi_host_interface_reset() - Issue host interface reset
nhi_reset_interface()
> + * @tb: Domain whose host interface is reset
This should take nhi as paramter.
> + *
> + * Resets the host interface by setting the RST bit in the host interface
> + * reset register. This brings the registers in the memory BAR to their
> + * default state and clears the End-to-End Flow Control state. Does nothing
> + * unless the host interface is known to need this (QUIRK_HOST_INTERFACE_RESET).
> + *
> + * The control channel is stopped over the reset because the reset clears
> + * the ring state as well.
> + */
This can just reset the NHI, e.g not look into the quirks.
> +void nhi_host_interface_reset(struct tb *tb)
> +{
> + struct tb_nhi *nhi = tb->nhi;
> + u32 val;
> +
> + if (!(nhi->quirks & QUIRK_HOST_INTERFACE_RESET))
> + return;
> +
> + val = ioread32(nhi->iobase + REG_CAPS);
> + /* Only v1 host interfaces implement the reset */
> + if (FIELD_GET(REG_CAPS_VERSION_MASK, val) >= REG_CAPS_VERSION_2)
> + return;
> +
> + dev_dbg(nhi->dev, "issuing host interface reset\n");
> +
> + tb_ctl_stop(tb->ctl);
But don't do these here. Just the actual reset. So that we can call it
whenever we want to do host internface reset.
> +
> + iowrite32(REG_HOST_INTERFACE_RESET_RST,
> + nhi->iobase + REG_HOST_INTERFACE_RESET);
> + /* Wait for tHIReset (10 ms) to complete */
> + usleep_range(10000, 20000);
> +
> + tb_ctl_start(tb->ctl);
> +}
> +
> static struct tb *nhi_select_cm(struct tb_nhi *nhi)
> {
> struct tb *tb;
> diff --git a/drivers/thunderbolt/nhi.h b/drivers/thunderbolt/nhi.h
> index d488eadadfce..9be0372b837e 100644
> --- a/drivers/thunderbolt/nhi.h
> +++ b/drivers/thunderbolt/nhi.h
> @@ -36,6 +36,7 @@ irqreturn_t nhi_msi(int irq, void *data);
> irqreturn_t ring_msix(int irq, void *data);
> int nhi_probe(struct tb_nhi *nhi);
> void nhi_shutdown(struct tb_nhi *nhi);
> +void nhi_host_interface_reset(struct tb *tb);
> extern const struct dev_pm_ops nhi_pm_ops;
>
> /**
> @@ -121,6 +122,7 @@ struct tb_nhi_ops {
> /* Host interface quirks */
> #define QUIRK_AUTO_CLEAR_INT BIT(0)
> #define QUIRK_E2E BIT(1)
> +#define QUIRK_HOST_INTERFACE_RESET BIT(2)
Let's call it QUIRK_RESET_DMA_ON_TEARDOWN or something like that.
> /*
> * Minimal number of vectors when we use MSI-X. Two for control channel
> diff --git a/drivers/thunderbolt/nhi_regs.h b/drivers/thunderbolt/nhi_regs.h
> index d6a197fabc74..99df60b6db36 100644
> --- a/drivers/thunderbolt/nhi_regs.h
> +++ b/drivers/thunderbolt/nhi_regs.h
> @@ -115,6 +115,10 @@ struct ring_desc {
> #define REG_CAPS_VERSION_MASK GENMASK(23, 16)
> #define REG_CAPS_VERSION_2 0x40
>
> +/* Host Interface Reset - resets TX/RX rings and E2E flow control counters */
> +#define REG_HOST_INTERFACE_RESET 0x39858
> +#define REG_HOST_INTERFACE_RESET_RST BIT(0)
> +
> #define REG_DMA_MISC 0x39864
> #define REG_DMA_MISC_INT_AUTO_CLEAR BIT(2)
> #define REG_DMA_MISC_DISABLE_AUTO_CLEAR BIT(17)
> diff --git a/drivers/thunderbolt/pci.c b/drivers/thunderbolt/pci.c
> index 8462ccb59b7e..02534287cb43 100644
> --- a/drivers/thunderbolt/pci.c
> +++ b/drivers/thunderbolt/pci.c
> @@ -62,6 +62,27 @@ static void nhi_pci_check_quirks(struct tb_nhi_pci *nhi_pci)
> nhi->quirks |= QUIRK_E2E;
> break;
> }
> + } else if (pdev->vendor == PCI_VENDOR_ID_AMD) {
> + switch (pdev->device) {
> + case 0x1120:
> + case 0x1121:
> + case 0x113b:
> + case 0x113c:
> + case 0x1155:
> + case 0x1158:
> + case 0x1159:
> + case 0x151c:
> + case 0x151d:
> + case 0x158d:
> + case 0x158e:
I would prefer if these are defined in nhi.h similarly what we do with the
Intel stuff so we have symbolic names for them.
> + /*
> + * These AMD hosts may hang the Tx ring when the
> + * DMA paths are torn down so they need the host
> + * interface reset after each teardown.
> + */
> + nhi->quirks |= QUIRK_HOST_INTERFACE_RESET;
> + break;
> + }
> }
> }
>
> diff --git a/drivers/thunderbolt/tb.c b/drivers/thunderbolt/tb.c
> index 47753a5c0f2e..f07f2ada2346 100644
> --- a/drivers/thunderbolt/tb.c
> +++ b/drivers/thunderbolt/tb.c
> @@ -2395,6 +2395,12 @@ static void __tb_disconnect_xdomain_paths(struct tb *tb, struct tb_xdomain *xd,
> * the same host router USB4 downstream port.
> */
> tb_enable_clx(sw);
> +
> + /*
> + * Some hosts may hang the Tx ring after the DMA paths are torn
> + * down so reset the host interface to prevent that.
> + */
> + nhi_host_interface_reset(tb);
Instead of here, do this in domain.c::tb_domain_disconnect_xdomain_paths()
and there you can also stop/start the control channel around this. You may
need to take the tb->lock here too so maybe add a helper.
I wanted to make tb.c (and domain.c) not call anything from the "upper"
layer (nhi.c) but I can't think of reasonable way to do it here, and there
is no way to handle this from nhi.c because it has no knowledge of XDomain
tunnels. Perhaps through indirection, nhi->ops->reset_interface?
> }
>
> static int tb_disconnect_xdomain_paths(struct tb *tb, struct tb_xdomain *xd,
> --
> 2.34.1
prev parent reply other threads:[~2026-08-05 12:28 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-05 11:18 [PATCH -next v2] thunderbolt: Add quirk to reset host interface on DMA path teardown for AMD USB4 routers Basavaraj Natikar
2026-08-05 12:28 ` Mika Westerberg [this message]
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=20260805122832.GH235112@black.igk.intel.com \
--to=mika.westerberg@linux.intel.com \
--cc=Basavaraj.Natikar@amd.com \
--cc=Mario.Limonciello@amd.com \
--cc=Sanath.S@amd.com \
--cc=YehezkelShB@gmail.com \
--cc=andreas.noever@gmail.com \
--cc=linux-usb@vger.kernel.org \
--cc=westeri@kernel.org \
/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