2024-05-22 01:28:19

by Daehwan Jung

[permalink] [raw]
Subject: [RFC] usb: host: xhci-mem: Write high first on erst base of secondary interrupter

ERSTBA_HI should be written first on secondary interrupter.
That's why secondary interrupter could be set while Host Controller
is already running.

[Synopsys]- The host controller was design to support ERST setting
during the RUN state. But since there is a limitation in controller
in supporting separate ERSTBA_HI and ERSTBA_LO programming,
It is supported when the ERSTBA is programmed in 64bit,
or in 32 bit mode ERSTBA_HI before ERSTBA_LO

[Synopsys]- The internal initialization of event ring fetches
the "Event Ring Segment Table Entry" based on the indication of
ERSTBA_LO written.

Signed-off-by: Daehwan Jung <[email protected]>
---
drivers/usb/host/xhci-mem.c | 5 ++++-
drivers/usb/host/xhci.h | 6 ++++++
2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
index 3100219..36ee704 100644
--- a/drivers/usb/host/xhci-mem.c
+++ b/drivers/usb/host/xhci-mem.c
@@ -2325,7 +2325,10 @@ xhci_add_interrupter(struct xhci_hcd *xhci, struct xhci_interrupter *ir,
erst_base = xhci_read_64(xhci, &ir->ir_set->erst_base);
erst_base &= ERST_BASE_RSVDP;
erst_base |= ir->erst.erst_dma_addr & ~ERST_BASE_RSVDP;
- xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
+ if (intr_num == 0)
+ xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
+ else
+ xhci_write_64_r(xhci, erst_base, &ir->ir_set->erst_base);

/* Set the event ring dequeue address of this interrupter */
xhci_set_hc_event_deq(xhci, ir);
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index 3041515..7951c0e 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -17,6 +17,7 @@
#include <linux/kernel.h>
#include <linux/usb/hcd.h>
#include <linux/io-64-nonatomic-lo-hi.h>
+#include <linux/io-64-nonatomic-hi-lo.h>

/* Code sharing between pci-quirks and xhci hcd */
#include "xhci-ext-caps.h"
@@ -1747,6 +1748,11 @@ static inline void xhci_write_64(struct xhci_hcd *xhci,
{
lo_hi_writeq(val, regs);
}
+static inline void xhci_write_64_r(struct xhci_hcd *xhci,
+ const u64 val, __le64 __iomem *regs)
+{
+ hi_lo_writeq(val, regs);
+}

static inline int xhci_link_trb_quirk(struct xhci_hcd *xhci)
{
--
2.7.4



2024-05-22 13:39:16

by Mathias Nyman

[permalink] [raw]
Subject: Re: [RFC] usb: host: xhci-mem: Write high first on erst base of secondary interrupter

On 22.5.2024 4.03, Daehwan Jung wrote:
> ERSTBA_HI should be written first on secondary interrupter.
> That's why secondary interrupter could be set while Host Controller
> is already running.
>
> [Synopsys]- The host controller was design to support ERST setting
> during the RUN state. But since there is a limitation in controller
> in supporting separate ERSTBA_HI and ERSTBA_LO programming,
> It is supported when the ERSTBA is programmed in 64bit,
> or in 32 bit mode ERSTBA_HI before ERSTBA_LO

xHCI specification 5.1 "Register Conventions "states that 64 bit
registers should be written in low-high order

>
> [Synopsys]- The internal initialization of event ring fetches
> the "Event Ring Segment Table Entry" based on the indication of
> ERSTBA_LO written.
>

Any idea if this is a common issue with this host?
Should other 64 bit registers also be written in reverse order.

> Signed-off-by: Daehwan Jung <[email protected]>
> ---
> drivers/usb/host/xhci-mem.c | 5 ++++-
> drivers/usb/host/xhci.h | 6 ++++++
> 2 files changed, 10 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
> index 3100219..36ee704 100644
> --- a/drivers/usb/host/xhci-mem.c
> +++ b/drivers/usb/host/xhci-mem.c
> @@ -2325,7 +2325,10 @@ xhci_add_interrupter(struct xhci_hcd *xhci, struct xhci_interrupter *ir,
> erst_base = xhci_read_64(xhci, &ir->ir_set->erst_base);
> erst_base &= ERST_BASE_RSVDP;
> erst_base |= ir->erst.erst_dma_addr & ~ERST_BASE_RSVDP;
> - xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
> + if (intr_num == 0)
> + xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
> + else
> + xhci_write_64_r(xhci, erst_base, &ir->ir_set->erst_base);

This may cause issues with other hosts expecting low-high order as stated
in the specification.

If all 64 bit registers should be written in high-low order for this host then
maybe set a quirk flag and change xhci_write_64()instead.

xhci_write_64(...)
{
if (xhci->quirks & XHCI_WRITE_64_HI_LO)
hi_lo_writeq(val, regs);
else
lo_hi_writeq(val, regs);
}


Thanks
Mathias

2024-05-23 13:37:04

by Mathias Nyman

[permalink] [raw]
Subject: Re: [RFC] usb: host: xhci-mem: Write high first on erst base of secondary interrupter

On 23.5.2024 7.43, Jung Daehwan wrote:
> On Wed, May 22, 2024 at 04:40:56PM +0300, Mathias Nyman wrote:
>> On 22.5.2024 4.03, Daehwan Jung wrote:
>>> ERSTBA_HI should be written first on secondary interrupter.
>>> That's why secondary interrupter could be set while Host Controller
>>> is already running.
>>>
>>> [Synopsys]- The host controller was design to support ERST setting
>>> during the RUN state. But since there is a limitation in controller
>>> in supporting separate ERSTBA_HI and ERSTBA_LO programming,
>>> It is supported when the ERSTBA is programmed in 64bit,
>>> or in 32 bit mode ERSTBA_HI before ERSTBA_LO
>>
>> xHCI specification 5.1 "Register Conventions "states that 64 bit
>> registers should be written in low-high order
>>
>>>
>>> [Synopsys]- The internal initialization of event ring fetches
>>> the "Event Ring Segment Table Entry" based on the indication of
>>> ERSTBA_LO written.
>>>
>>
>> Any idea if this is a common issue with this host?
>> Should other 64 bit registers also be written in reverse order.
>>
>>> Signed-off-by: Daehwan Jung <[email protected]>
>>> ---
>>> drivers/usb/host/xhci-mem.c | 5 ++++-
>>> drivers/usb/host/xhci.h | 6 ++++++
>>> 2 files changed, 10 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/usb/host/xhci-mem.c b/drivers/usb/host/xhci-mem.c
>>> index 3100219..36ee704 100644
>>> --- a/drivers/usb/host/xhci-mem.c
>>> +++ b/drivers/usb/host/xhci-mem.c
>>> @@ -2325,7 +2325,10 @@ xhci_add_interrupter(struct xhci_hcd *xhci, struct xhci_interrupter *ir,
>>> erst_base = xhci_read_64(xhci, &ir->ir_set->erst_base);
>>> erst_base &= ERST_BASE_RSVDP;
>>> erst_base |= ir->erst.erst_dma_addr & ~ERST_BASE_RSVDP;
>>> - xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
>>> + if (intr_num == 0)
>>> + xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
>>> + else
>>> + xhci_write_64_r(xhci, erst_base, &ir->ir_set->erst_base);
>>
>> This may cause issues with other hosts expecting low-high order as stated
>> in the specification.
>>
>> If all 64 bit registers should be written in high-low order for this host then
>> maybe set a quirk flag and change xhci_write_64()instead.
>>
>> xhci_write_64(...)
>> {
>> if (xhci->quirks & XHCI_WRITE_64_HI_LO)
>> hi_lo_writeq(val, regs);
>> else
>> lo_hi_writeq(val, regs);
>> }
>>
>
> Mathias, Thanks for the comment.
>
> I've seen this issue only writing the base address of ERST.
> It's better to use a quirk flag as you said.
> How about using the quirk only in xhci_add_interrupter?
>
> @@ -2325,7 +2325,10 @@ xhci_add_interrupter(struct xhci_hcd *xhci, struct xhci_interrupter *ir,
> erst_base = xhci_read_64(xhci, &ir->ir_set->erst_base);
> erst_base &= ERST_BASE_RSVDP;
> erst_base |= ir->erst.erst_dma_addr & ~ERST_BASE_RSVDP;
> xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
> if (xhci->quirks & XHCI_WRITE_64_HI_LO)
> xhci_write_64_r(xhci, erst_base, &ir->ir_set->erst_base);
> else
> xhci_write_64(xhci, erst_base, &ir->ir_set->erst_base);
>

This works.
Maybe even skip the xhci_write_64_r() helper and just use hi_lo_writeq() directly.

Thanks
Mathias