| Message ID | 20191014141718.22603-2-narmstrong@baylibre.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | arm64: g12-common: parkmode_disable_ss_quirk on DWC3 controller | expand |
On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: > This patch updates the documentation with the information related > to the quirks that needs to be added for disabling all SuperSpeed XHCi > instances in park mode. > > CC: Dongjin Kim <tobetter@gmail.com> > Cc: Jianxin Pan <jianxin.pan@amlogic.com> > Reported-by: Tim <elatllat@gmail.com> > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > --- > Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ > 1 file changed, 2 insertions(+) Sigh, what's one more to the never ending list of quirks... Acked-by: Rob Herring <robh@kernel.org>
Thanks for working on this Neil,
Is there something that needs doing for this patch to make it into 5.3 or 5.4?
As previously mentioned the patch set fixes the issue on affected hardware;
https://patchwork.kernel.org/patch/11164515/
On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote:
>
> On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote:
> > This patch updates the documentation with the information related
> > to the quirks that needs to be added for disabling all SuperSpeed XHCi
> > instances in park mode.
> >
> > CC: Dongjin Kim <tobetter@gmail.com>
> > Cc: Jianxin Pan <jianxin.pan@amlogic.com>
> > Reported-by: Tim <elatllat@gmail.com>
> > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
> > ---
> > Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++
> > 1 file changed, 2 insertions(+)
>
> Sigh, what's one more to the never ending list of quirks...
>
> Acked-by: Rob Herring <robh@kernel.org>
Hi Neil As I got the information from Synopsys, this bug exists on current IP versions, and per my tests with external USB3 hub + 2 Super speed udisks on data read by dd, I can reproduce this issue with different kernel versions, also I didn't see obvious performance drop by dd tests after disable park mode for super speed, so should we just disable it by default so no need a quirk? Li Jun Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道： > > Thanks for working on this Neil, > Is there something that needs doing for this patch to make it into 5.3 or 5.4? > As previously mentioned the patch set fixes the issue on affected hardware; > https://patchwork.kernel.org/patch/11164515/ > > > > On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote: > > > > On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: > > > This patch updates the documentation with the information related > > > to the quirks that needs to be added for disabling all SuperSpeed XHCi > > > instances in park mode. > > > > > > CC: Dongjin Kim <tobetter@gmail.com> > > > Cc: Jianxin Pan <jianxin.pan@amlogic.com> > > > Reported-by: Tim <elatllat@gmail.com> > > > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > > > --- > > > Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ > > > 1 file changed, 2 insertions(+) > > > > Sigh, what's one more to the never ending list of quirks... > > > > Acked-by: Rob Herring <robh@kernel.org>
Speculation; Maybe the kernel maintainers prefer to optimistically permit future products to easily remove workarounds via quirk flags. Even if data from testing were shown, and it did not impact performance, code reduction and clarity are desirable. On Sun, Nov 10, 2019 at 8:58 PM Jun Li <lijun.kernel@gmail.com> wrote: > > Hi Neil > > As I got the information from Synopsys, this bug exists on current IP versions, > and per my tests with external USB3 hub + 2 Super speed udisks on data > read by dd, I can reproduce this issue with different kernel versions, also I > didn't see obvious performance drop by dd tests after disable park mode for > super speed, so should we just disable it by default so no need a quirk? > > Li Jun > > Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道： > > > > Thanks for working on this Neil, > > Is there something that needs doing for this patch to make it into 5.3 or 5.4? > > As previously mentioned the patch set fixes the issue on affected hardware; > > https://patchwork.kernel.org/patch/11164515/ > > > > > > > > On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote: > > > > > > On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: > > > > This patch updates the documentation with the information related > > > > to the quirks that needs to be added for disabling all SuperSpeed XHCi > > > > instances in park mode. > > > > > > > > CC: Dongjin Kim <tobetter@gmail.com> > > > > Cc: Jianxin Pan <jianxin.pan@amlogic.com> > > > > Reported-by: Tim <elatllat@gmail.com> > > > > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > > > > --- > > > > Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ > > > > 1 file changed, 2 insertions(+) > > > > > > Sigh, what's one more to the never ending list of quirks... > > > > > > Acked-by: Rob Herring <robh@kernel.org>
by a quirk, those platforms without this quirk can't benefit this fix.
if later this bug is fixed by some reversion of dwc3 IP, we can easily
remove the workaround by a IP version check:
if (dwc->revision < DWC3_REVISION_PARK_MODE_FIXED)
{
disable park mode for ss;
}
Li Jun
Tim <elatllat@gmail.com> 于2019年11月11日周一 下午10:05写道：
>
> Speculation;
>
> Maybe the kernel maintainers prefer to optimistically permit future
> products to easily remove workarounds via quirk flags.
> Even if data from testing were shown, and it did not impact
> performance, code reduction and clarity are desirable.
>
> On Sun, Nov 10, 2019 at 8:58 PM Jun Li <lijun.kernel@gmail.com> wrote:
> >
> > Hi Neil
> >
> > As I got the information from Synopsys, this bug exists on current IP versions,
> > and per my tests with external USB3 hub + 2 Super speed udisks on data
> > read by dd, I can reproduce this issue with different kernel versions, also I
> > didn't see obvious performance drop by dd tests after disable park mode for
> > super speed, so should we just disable it by default so no need a quirk?
> >
> > Li Jun
> >
> > Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道：
> > >
> > > Thanks for working on this Neil,
> > > Is there something that needs doing for this patch to make it into 5.3 or 5.4?
> > > As previously mentioned the patch set fixes the issue on affected hardware;
> > > https://patchwork.kernel.org/patch/11164515/
> > >
> > >
> > >
> > > On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote:
> > > >
> > > > On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote:
> > > > > This patch updates the documentation with the information related
> > > > > to the quirks that needs to be added for disabling all SuperSpeed XHCi
> > > > > instances in park mode.
> > > > >
> > > > > CC: Dongjin Kim <tobetter@gmail.com>
> > > > > Cc: Jianxin Pan <jianxin.pan@amlogic.com>
> > > > > Reported-by: Tim <elatllat@gmail.com>
> > > > > Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
> > > > > ---
> > > > > Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++
> > > > > 1 file changed, 2 insertions(+)
> > > >
> > > > Sigh, what's one more to the never ending list of quirks...
> > > >
> > > > Acked-by: Rob Herring <robh@kernel.org>
Hi Li, On 11/11/2019 02:58, Jun Li wrote: > Hi Neil > > As I got the information from Synopsys, this bug exists on current IP versions, > and per my tests with external USB3 hub + 2 Super speed udisks on data > read by dd, I can reproduce this issue with different kernel versions, also I > didn't see obvious performance drop by dd tests after disable park mode for > super speed, so should we just disable it by default so no need a quirk? I don't have any opinion on this, I think the USB & DWC3 maintainers should decide how to handle this. Did Synopsys specified a range of affected IP version ? Neil > > Li Jun > > Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道： >> >> Thanks for working on this Neil, >> Is there something that needs doing for this patch to make it into 5.3 or 5.4? >> As previously mentioned the patch set fixes the issue on affected hardware; >> https://patchwork.kernel.org/patch/11164515/ >> >> >> >> On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote: >>> >>> On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: >>>> This patch updates the documentation with the information related >>>> to the quirks that needs to be added for disabling all SuperSpeed XHCi >>>> instances in park mode. >>>> >>>> CC: Dongjin Kim <tobetter@gmail.com> >>>> Cc: Jianxin Pan <jianxin.pan@amlogic.com> >>>> Reported-by: Tim <elatllat@gmail.com> >>>> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> >>>> --- >>>> Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ >>>> 1 file changed, 2 insertions(+) >>> >>> Sigh, what's one more to the never ending list of quirks... >>> >>> Acked-by: Rob Herring <robh@kernel.org>
This bug exists on all current versions per information I got from Synopsys. + Thinh Nguyen <thinhn@synopsys.com>. Li Jun Neil Armstrong <narmstrong@baylibre.com> 于2019年11月12日周二 下午5:00写道： > > Hi Li, > > On 11/11/2019 02:58, Jun Li wrote: > > Hi Neil > > > > As I got the information from Synopsys, this bug exists on current IP versions, > > and per my tests with external USB3 hub + 2 Super speed udisks on data > > read by dd, I can reproduce this issue with different kernel versions, also I > > didn't see obvious performance drop by dd tests after disable park mode for > > super speed, so should we just disable it by default so no need a quirk? > > I don't have any opinion on this, I think the USB & DWC3 maintainers should decide > how to handle this. > > Did Synopsys specified a range of affected IP version ? > > Neil > > > > > Li Jun > > > > Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道： > >> > >> Thanks for working on this Neil, > >> Is there something that needs doing for this patch to make it into 5.3 or 5.4? > >> As previously mentioned the patch set fixes the issue on affected hardware; > >> https://patchwork.kernel.org/patch/11164515/ > >> > >> > >> > >> On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote: > >>> > >>> On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: > >>>> This patch updates the documentation with the information related > >>>> to the quirks that needs to be added for disabling all SuperSpeed XHCi > >>>> instances in park mode. > >>>> > >>>> CC: Dongjin Kim <tobetter@gmail.com> > >>>> Cc: Jianxin Pan <jianxin.pan@amlogic.com> > >>>> Reported-by: Tim <elatllat@gmail.com> > >>>> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> > >>>> --- > >>>> Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ > >>>> 1 file changed, 2 insertions(+) > >>> > >>> Sigh, what's one more to the never ending list of quirks... > >>> > >>> Acked-by: Rob Herring <robh@kernel.org> >
Hi, Jun Li wrote: > This bug exists on all current versions per information I got from Synopsys. > + Thinh Nguyen <thinhn@synopsys.com>. > > Li Jun The GUCTL1.PARKMODE_DISABLE_SS is only available in dwc_usb3 controller running in host mode. This should not be set for other IPs. This can be disabled by default based on IP, but I recommend to have a property to enable this feature for devices that need this. > > Neil Armstrong <narmstrong@baylibre.com> 于2019年11月12日周二 下午5:00写道： >> Hi Li, >> >> On 11/11/2019 02:58, Jun Li wrote: >>> Hi Neil >>> >>> As I got the information from Synopsys, this bug exists on current IP versions, >>> and per my tests with external USB3 hub + 2 Super speed udisks on data >>> read by dd, I can reproduce this issue with different kernel versions, also I >>> didn't see obvious performance drop by dd tests after disable park mode for >>> super speed, so should we just disable it by default so no need a quirk? >> I don't have any opinion on this, I think the USB & DWC3 maintainers should decide >> how to handle this. >> >> Did Synopsys specified a range of affected IP version ? >> >> Neil >> >>> Li Jun >>> >>> Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道： >>>> Thanks for working on this Neil, >>>> Is there something that needs doing for this patch to make it into 5.3 or 5.4? >>>> As previously mentioned the patch set fixes the issue on affected hardware; >>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__patchwork.kernel.org_patch_11164515_&d=DwIFaQ&c=DPL6_X_6JkXFx7AXWqB0tg&r=u9FYoxKtyhjrGFcyixFYqTjw1ZX0VsG2d8FCmzkTY-w&m=BCfRHM9e86039Su0-QvK_yUjNnYrnl_BcvxTtViypT8&s=aV5u7uxNX6oTLB3_-qNuzrNvyiqx_rPC6P9TMqG-LNE&e= >>>> >>>> >>>> >>>> On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote: >>>>> On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: >>>>>> This patch updates the documentation with the information related >>>>>> to the quirks that needs to be added for disabling all SuperSpeed XHCi >>>>>> instances in park mode. >>>>>> >>>>>> CC: Dongjin Kim <tobetter@gmail.com> >>>>>> Cc: Jianxin Pan <jianxin.pan@amlogic.com> >>>>>> Reported-by: Tim <elatllat@gmail.com> >>>>>> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> >>>>>> --- >>>>>> Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ >>>>>> 1 file changed, 2 insertions(+) >>>>> Sigh, what's one more to the never ending list of quirks... >>>>> >>>>> Acked-by: Rob Herring <robh@kernel.org> BR, Thinh
Hi, On 12/11/2019 21:03, Thinh Nguyen wrote: > Hi, > > Jun Li wrote: >> This bug exists on all current versions per information I got from Synopsys. >> + Thinh Nguyen <thinhn@synopsys.com>. >> >> Li Jun > > The GUCTL1.PARKMODE_DISABLE_SS is only available in dwc_usb3 controller > running in host mode. This should not be set for other IPs. > > This can be disabled by default based on IP, but I recommend to have a > property to enable this feature for devices that need this. Thanks, so this patchset is the right way to solve this. Felipe, would it be possible to have it merged for next release ? Should I resend it ? Thanks, Neil > > >> >> Neil Armstrong <narmstrong@baylibre.com> 于2019年11月12日周二 下午5:00写道： >>> Hi Li, >>> >>> On 11/11/2019 02:58, Jun Li wrote: >>>> Hi Neil >>>> >>>> As I got the information from Synopsys, this bug exists on current IP versions, >>>> and per my tests with external USB3 hub + 2 Super speed udisks on data >>>> read by dd, I can reproduce this issue with different kernel versions, also I >>>> didn't see obvious performance drop by dd tests after disable park mode for >>>> super speed, so should we just disable it by default so no need a quirk? >>> I don't have any opinion on this, I think the USB & DWC3 maintainers should decide >>> how to handle this. >>> >>> Did Synopsys specified a range of affected IP version ? >>> >>> Neil >>> >>>> Li Jun >>>> >>>> Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道： >>>>> Thanks for working on this Neil, >>>>> Is there something that needs doing for this patch to make it into 5.3 or 5.4? >>>>> As previously mentioned the patch set fixes the issue on affected hardware; >>>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__patchwork.kernel.org_patch_11164515_&d=DwIFaQ&c=DPL6_X_6JkXFx7AXWqB0tg&r=u9FYoxKtyhjrGFcyixFYqTjw1ZX0VsG2d8FCmzkTY-w&m=BCfRHM9e86039Su0-QvK_yUjNnYrnl_BcvxTtViypT8&s=aV5u7uxNX6oTLB3_-qNuzrNvyiqx_rPC6P9TMqG-LNE&e= >>>>> >>>>> >>>>> >>>>> On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote: >>>>>> On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote: >>>>>>> This patch updates the documentation with the information related >>>>>>> to the quirks that needs to be added for disabling all SuperSpeed XHCi >>>>>>> instances in park mode. >>>>>>> >>>>>>> CC: Dongjin Kim <tobetter@gmail.com> >>>>>>> Cc: Jianxin Pan <jianxin.pan@amlogic.com> >>>>>>> Reported-by: Tim <elatllat@gmail.com> >>>>>>> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> >>>>>>> --- >>>>>>> Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ >>>>>>> 1 file changed, 2 insertions(+) >>>>>> Sigh, what's one more to the never ending list of quirks... >>>>>> >>>>>> Acked-by: Rob Herring <robh@kernel.org> > > BR, > Thinh >
Just writing to say;
this patch is still performing well on 5.5.0-rc1, and
there is no speed improvement by adding DWC3_GUCTL_NAK_PER_ENH_HS BIT(17)
I get up to 3890 ms second lag with "ping -s 5120 $IP" when
copying between USB HDDs either way, but at least this patch stops the
crashing.
On Tue, Dec 10, 2019 at 3:55 AM Neil Armstrong <narmstrong@baylibre.com> wrote:
>
> Hi,
>
> On 12/11/2019 21:03, Thinh Nguyen wrote:
> > Hi,
> >
> > Jun Li wrote:
> >> This bug exists on all current versions per information I got from Synopsys.
> >> + Thinh Nguyen <thinhn@synopsys.com>.
> >>
> >> Li Jun
> >
> > The GUCTL1.PARKMODE_DISABLE_SS is only available in dwc_usb3 controller
> > running in host mode. This should not be set for other IPs.
> >
> > This can be disabled by default based on IP, but I recommend to have a
> > property to enable this feature for devices that need this.
>
> Thanks, so this patchset is the right way to solve this.
>
> Felipe, would it be possible to have it merged for next release ?
> Should I resend it ?
>
> Thanks,
> Neil
>
> >
> >
> >>
> >> Neil Armstrong <narmstrong@baylibre.com> 于2019年11月12日周二 下午5:00写道：
> >>> Hi Li,
> >>>
> >>> On 11/11/2019 02:58, Jun Li wrote:
> >>>> Hi Neil
> >>>>
> >>>> As I got the information from Synopsys, this bug exists on current IP versions,
> >>>> and per my tests with external USB3 hub + 2 Super speed udisks on data
> >>>> read by dd, I can reproduce this issue with different kernel versions, also I
> >>>> didn't see obvious performance drop by dd tests after disable park mode for
> >>>> super speed, so should we just disable it by default so no need a quirk?
> >>> I don't have any opinion on this, I think the USB & DWC3 maintainers should decide
> >>> how to handle this.
> >>>
> >>> Did Synopsys specified a range of affected IP version ?
> >>>
> >>> Neil
> >>>
> >>>> Li Jun
> >>>>
> >>>> Tim <elatllat@gmail.com> 于2019年11月11日周一 上午8:42写道：
> >>>>> Thanks for working on this Neil,
> >>>>> Is there something that needs doing for this patch to make it into 5.3 or 5.4?
> >>>>> As previously mentioned the patch set fixes the issue on affected hardware;
> >>>>> https://urldefense.proofpoint.com/v2/url?u=https-3A__patchwork.kernel.org_patch_11164515_&d=DwIFaQ&c=DPL6_X_6JkXFx7AXWqB0tg&r=u9FYoxKtyhjrGFcyixFYqTjw1ZX0VsG2d8FCmzkTY-w&m=BCfRHM9e86039Su0-QvK_yUjNnYrnl_BcvxTtViypT8&s=aV5u7uxNX6oTLB3_-qNuzrNvyiqx_rPC6P9TMqG-LNE&e=
> >>>>>
> >>>>>
> >>>>>
> >>>>> On Wed, Oct 23, 2019 at 4:11 PM Rob Herring <robh@kernel.org> wrote:
> >>>>>> On Mon, Oct 14, 2019 at 04:17:16PM +0200, Neil Armstrong wrote:
> >>>>>>> This patch updates the documentation with the information related
> >>>>>>> to the quirks that needs to be added for disabling all SuperSpeed XHCi
> >>>>>>> instances in park mode.
> >>>>>>>
> >>>>>>> CC: Dongjin Kim <tobetter@gmail.com>
> >>>>>>> Cc: Jianxin Pan <jianxin.pan@amlogic.com>
> >>>>>>> Reported-by: Tim <elatllat@gmail.com>
> >>>>>>> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
> >>>>>>> ---
> >>>>>>> Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++
> >>>>>>> 1 file changed, 2 insertions(+)
> >>>>>> Sigh, what's one more to the never ending list of quirks...
> >>>>>>
> >>>>>> Acked-by: Rob Herring <robh@kernel.org>
> >
> > BR,
> > Thinh
> >
>
This patch is still good on 5.4 and 5.5, and the following network lag was solved by separating the USB and NIC smp_affinity_list. On Tue, Dec 10, 2019, 7:45 PM Tim <elatllat@gmail.com> wrote: > > Just writing to say; > this patch is still performing well on 5.5.0-rc1, and > there is no speed improvement by adding DWC3_GUCTL_NAK_PER_ENH_HS BIT(17) > I get up to 3890 ms second lag with "ping -s 5120 $IP" when > copying between USB HDDs either way, but at least this patch stops the > crashing.
diff --git a/Documentation/devicetree/bindings/usb/dwc3.txt b/Documentation/devicetree/bindings/usb/dwc3.txt index 66780a47ad85..c977a3ba2f35 100644 --- a/Documentation/devicetree/bindings/usb/dwc3.txt +++ b/Documentation/devicetree/bindings/usb/dwc3.txt @@ -75,6 +75,8 @@ Optional properties: from P0 to P1/P2/P3 without delay. - snps,dis-tx-ipgap-linecheck-quirk: when set, disable u2mac linestate check during HS transmit. + - snps,parkmode-disable-ss-quirk: when set, all SuperSpeed bus instances in + park mode are disabled. - snps,dis_metastability_quirk: when set, disable metastability workaround. CAUTION: use only if you are absolutely sure of it. - snps,is-utmi-l1-suspend: true when DWC3 asserts output signal
This patch updates the documentation with the information related to the quirks that needs to be added for disabling all SuperSpeed XHCi instances in park mode. CC: Dongjin Kim <tobetter@gmail.com> Cc: Jianxin Pan <jianxin.pan@amlogic.com> Reported-by: Tim <elatllat@gmail.com> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> --- Documentation/devicetree/bindings/usb/dwc3.txt | 2 ++ 1 file changed, 2 insertions(+)