linux-kernel - Re: [PATCH 2/2] of: property: fw_devlink: Set 'optional_con_dev' for parse_power

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 31 Aug 2021 10:50:32 -0700
From:   Saravana Kannan <saravanak@...gle.com>
To:     Ulf Hansson <ulf.hansson@...aro.org>
Cc:     Rob Herring <robh@...nel.org>, devicetree@...r.kernel.org,
        "Rafael J . Wysocki" <rjw@...ysocki.net>,
        Stephen Boyd <sboyd@...nel.org>,
        Dmitry Osipenko <digetx@...il.com>, linux-pm@...r.kernel.org,
        linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH 2/2] of: property: fw_devlink: Set 'optional_con_dev' for parse_power_domains

On Tue, Aug 31, 2021 at 3:21 AM Ulf Hansson <ulf.hansson@...aro.org> wrote:
>
> The power-domain DT bindings [1] doesn't enforce a compatible string for a
> provider node, even if this is common to use. In particular, when
> describing a hierarchy with parent/child power-domains, as the psci DT
> bindings [2] for example, it's sometimes not applicable to use a compatible
> string.

Ok, and fw_devlink handles that -- provider not having a compatible
string is pretty common. In these cases, the parent node is the actual
device that gets probed and registers the provider. So fw_devlink will
create a link from the consumer to the parent device node.

> Therefore, let's set the 'optional_con_dev' to true to avoid creating
> incorrect fw_devlinks for power-domains.

This part doesn't make sense or is incomplete. What is being done incorrectly?

>
> [1] Documentation/devicetree/bindings/power/power-domain.yaml
> [2] Documentation/devicetree/bindings/arm/psci.yaml
>
> Signed-off-by: Ulf Hansson <ulf.hansson@...aro.org>
> ---
>
> Some more details of what goes on here. I have added a debug print in
> of_link_to_phandle() to see the fw_devlinks that gets created.
>
> This is what happens on Dragonboard 410c when 'optional_con_dev' isn't set:
> ...
> [    0.041274] device: 'psci': device_add
> [    0.041366] OF: Linking power-domain-cpu0 (consumer) to psci (supplier)
> [    0.041395] OF: Linking power-domain-cpu1 (consumer) to psci (supplier)
> [    0.041423] OF: Linking power-domain-cpu2 (consumer) to psci (supplier)
> [    0.041451] OF: Linking power-domain-cpu3 (consumer) to psci (supplier)
> [    0.041494] device: 'platform:psci--platform:psci': device_add
> [    0.041556] platform psci: Linked as a sync state only consumer to psci
> ...
>
> This is what happens on Dragonboard 410c when 'optional_con_dev' is set:
> ...
> [    0.041179] device: 'psci': device_add
> [    0.041265] OF: Not linking psci to psci - is descendant
> [    0.041293] OF: Not linking psci to psci - is descendant
> [    0.041319] OF: Not linking psci to psci - is descendant
> [    0.041346] OF: Not linking psci to psci - is descendant
> ...

Can you please explain what exactly is going on that's wrong here? I
notice that psci is not probed as a device at all. And when you aren't
setting this flag the only difference I see is the creating of a sync
state only link -- which shouldn't matter here because you don't even
have a driver implemented.

> The relevant dtsi file:
> arch/arm64/boot/dts/qcom/msm8916.dtsi
>
> Kind regards
> Ulf Hansson
>
> ---
>  drivers/of/property.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/of/property.c b/drivers/of/property.c
> index 2babb1807228..4d607fdbea24 100644
> --- a/drivers/of/property.c
> +++ b/drivers/of/property.c
> @@ -1356,7 +1356,7 @@ static const struct supplier_bindings of_supplier_bindings[] = {
>         { .parse_prop = parse_io_channels, },
>         { .parse_prop = parse_interrupt_parent, },
>         { .parse_prop = parse_dmas, .optional = true, },
> -       { .parse_prop = parse_power_domains, },
> +       { .parse_prop = parse_power_domains, .optional_con_dev = true, },

This change is just shooting in dark/completely unrelated to the
commit text. This is just saying the actual consumer is a level up
from where the property is listed (eg: remote-endpoint). It just
happens to fix your case for unrelated reasons.

Definite Nak as this *will* break other cases.

-Saravana


>         { .parse_prop = parse_hwlocks, },
>         { .parse_prop = parse_extcon, },
>         { .parse_prop = parse_nvmem_cells, },
> --
> 2.25.1
>