Regression with "PCI: qcom: Add support for tx term offset for rev 2.1.0"

Regression with "PCI: qcom: Add support for tx term offset for rev 2.1.0"

[Ilia Mirkin]

Hi Ansuel,

I'm on an ifc6410 (APQ8064), and the latest v5.9-rc's hang during PCIe
init. I just get:

[    1.191861] qcom-pcie 1b500000.pci: host bridge /soc/pci@1b500000 ranges:
[    1.197756] qcom-pcie 1b500000.pci:       IO
0x000fe00000..0x000fefffff -> 0x0000000000
[    1.205625] qcom-pcie 1b500000.pci:      MEM
0x0008000000..0x000fdfffff -> 0x0008000000

and then it hangs forever. On a working kernel, the next message is e.g.

[    6.737388] qcom-pcie 1b500000.pci: Link up

A bisect led to

$ git bisect good
de3c4bf648975ea0b1d344d811e9b0748907b47c is the first bad commit
commit de3c4bf648975ea0b1d344d811e9b0748907b47c
Author: Ansuel Smith <ansuelsmth@gmail.com>
Date:   Mon Jun 15 23:06:04 2020 +0200

    PCI: qcom: Add support for tx term offset for rev 2.1.0

    Add tx term offset support to pcie qcom driver need in some revision of
    the ipq806x SoC. Ipq8064 needs tx term offset set to 7.

    Link: https://lore.kernel.org/r/20200615210608.21469-9-ansuelsmth@gmail.com
    Fixes: 82a823833f4e ("PCI: qcom: Add Qualcomm PCIe controller driver")
    Signed-off-by: Sham Muthayyan <smuthayy@codeaurora.org>
    Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
    Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
    Acked-by: Stanimir Varbanov <svarbanov@mm-sol.com>
    Cc: stable@vger.kernel.org # v4.5+

 drivers/pci/controller/dwc/pcie-qcom.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)

And indeed reverting this commit on top of v5.9-rc3 gets back to a
working system. I have no idea what PHY_REFCLK_USE_PAD is, but it
seems like clearing it is messing things up for me. (As everything
else seems like it should be identical for me.)

Let me know if you want me to test anything, or if the best path is to
just revert for now.

Cheers,

Ilia Mirkin
imirkin@alum.mit.edu

R: Regression with "PCI: qcom: Add support for tx term offset for rev 2.1.0"

[ansuelsmth]

> -----Messaggio originale-----
> Da: Ilia Mirkin <imirkin@alum.mit.edu>
> Inviato: lunedì 7 settembre 2020 00:59
> A: Ansuel Smith <ansuelsmth@gmail.com>
> Cc: linux-arm-msm@vger.kernel.org; Linux PCI <linux-pci@vger.kernel.org>
> Oggetto: Regression with "PCI: qcom: Add support for tx term offset for rev
> 2.1.0"
> 
> Hi Ansuel,
> 
> I'm on an ifc6410 (APQ8064), and the latest v5.9-rc's hang during PCIe
> init. I just get:
> 
> [    1.191861] qcom-pcie 1b500000.pci: host bridge /soc/pci@1b500000
> ranges:
> [    1.197756] qcom-pcie 1b500000.pci:       IO
> 0x000fe00000..0x000fefffff -> 0x0000000000
> [    1.205625] qcom-pcie 1b500000.pci:      MEM
> 0x0008000000..0x000fdfffff -> 0x0008000000
> 
> and then it hangs forever. On a working kernel, the next message is e.g.
> 
> [    6.737388] qcom-pcie 1b500000.pci: Link up
> 
> A bisect led to
> 
> $ git bisect good
> de3c4bf648975ea0b1d344d811e9b0748907b47c is the first bad commit
> commit de3c4bf648975ea0b1d344d811e9b0748907b47c
> Author: Ansuel Smith <ansuelsmth@gmail.com>
> Date:   Mon Jun 15 23:06:04 2020 +0200
> 
>     PCI: qcom: Add support for tx term offset for rev 2.1.0
> 
>     Add tx term offset support to pcie qcom driver need in some revision of
>     the ipq806x SoC. Ipq8064 needs tx term offset set to 7.
> 
>     Link: https://lore.kernel.org/r/20200615210608.21469-9-
> ansuelsmth@gmail.com
>     Fixes: 82a823833f4e ("PCI: qcom: Add Qualcomm PCIe controller driver")
>     Signed-off-by: Sham Muthayyan <smuthayy@codeaurora.org>
>     Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
>     Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>     Acked-by: Stanimir Varbanov <svarbanov@mm-sol.com>
>     Cc: stable@vger.kernel.org # v4.5+
> 
>  drivers/pci/controller/dwc/pcie-qcom.c | 17 ++++++++++++++++-
>  1 file changed, 16 insertions(+), 1 deletion(-)
> 
> And indeed reverting this commit on top of v5.9-rc3 gets back to a
> working system. I have no idea what PHY_REFCLK_USE_PAD is, but it
> seems like clearing it is messing things up for me. (As everything
> else seems like it should be identical for me.)
> 
> Let me know if you want me to test anything, or if the best path is to
> just revert for now.
> 
> Cheers,
> 

Thanks for the report. Can you confirm that by only removing 
PHY_REFCLK_USE_PAD the problem is fixed? Wonder if it's better to
just make a patch to restrict the padding to only ipq806x and backport that.
(or revert and repush a better patch). What do you think?

> Ilia Mirkin
> imirkin@alum.mit.edu

Re: Regression with "PCI: qcom: Add support for tx term offset for rev 2.1.0"

[Ilia Mirkin]

On Sun, Sep 6, 2020 at 8:01 PM <ansuelsmth@gmail.com> wrote:
>
>
>
> > -----Messaggio originale-----
> > Da: Ilia Mirkin <imirkin@alum.mit.edu>
> > Inviato: lunedì 7 settembre 2020 00:59
> > A: Ansuel Smith <ansuelsmth@gmail.com>
> > Cc: linux-arm-msm@vger.kernel.org; Linux PCI <linux-pci@vger.kernel.org>
> > Oggetto: Regression with "PCI: qcom: Add support for tx term offset for rev
> > 2.1.0"
> >
> > Hi Ansuel,
> >
> > I'm on an ifc6410 (APQ8064), and the latest v5.9-rc's hang during PCIe
> > init. I just get:
> >
> > [    1.191861] qcom-pcie 1b500000.pci: host bridge /soc/pci@1b500000
> > ranges:
> > [    1.197756] qcom-pcie 1b500000.pci:       IO
> > 0x000fe00000..0x000fefffff -> 0x0000000000
> > [    1.205625] qcom-pcie 1b500000.pci:      MEM
> > 0x0008000000..0x000fdfffff -> 0x0008000000
> >
> > and then it hangs forever. On a working kernel, the next message is e.g.
> >
> > [    6.737388] qcom-pcie 1b500000.pci: Link up
> >
> > A bisect led to
> >
> > $ git bisect good
> > de3c4bf648975ea0b1d344d811e9b0748907b47c is the first bad commit
> > commit de3c4bf648975ea0b1d344d811e9b0748907b47c
> > Author: Ansuel Smith <ansuelsmth@gmail.com>
> > Date:   Mon Jun 15 23:06:04 2020 +0200
> >
> >     PCI: qcom: Add support for tx term offset for rev 2.1.0
> >
> >     Add tx term offset support to pcie qcom driver need in some revision of
> >     the ipq806x SoC. Ipq8064 needs tx term offset set to 7.
> >
> >     Link: https://lore.kernel.org/r/20200615210608.21469-9-
> > ansuelsmth@gmail.com
> >     Fixes: 82a823833f4e ("PCI: qcom: Add Qualcomm PCIe controller driver")
> >     Signed-off-by: Sham Muthayyan <smuthayy@codeaurora.org>
> >     Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> >     Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> >     Acked-by: Stanimir Varbanov <svarbanov@mm-sol.com>
> >     Cc: stable@vger.kernel.org # v4.5+
> >
> >  drivers/pci/controller/dwc/pcie-qcom.c | 17 ++++++++++++++++-
> >  1 file changed, 16 insertions(+), 1 deletion(-)
> >
> > And indeed reverting this commit on top of v5.9-rc3 gets back to a
> > working system. I have no idea what PHY_REFCLK_USE_PAD is, but it
> > seems like clearing it is messing things up for me. (As everything
> > else seems like it should be identical for me.)
> >
> > Let me know if you want me to test anything, or if the best path is to
> > just revert for now.
> >
> > Cheers,
> >
>
> Thanks for the report. Can you confirm that by only removing
> PHY_REFCLK_USE_PAD the problem is fixed? Wonder if it's better to
> just make a patch to restrict the padding to only ipq806x and backport that.
> (or revert and repush a better patch). What do you think?

Confirmed. The following patch makes it work for me (as expected):

https://lore.kernel.org/linux-arm-msm/20200907011238.3401-1-imirkin@alum.mit.edu/

Feel free to follow up with any feedback there.

Cheers,

  -ilia