netdev - PHY reset may still be asserted during MDIO probe

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-next>] [day] [month] [year] [list]

Message-ID: <CAMuHMdXno2OUHqsAfO0z43JmGkFehD+FJ2dEjEsr_P53oAAPxA@mail.gmail.com>
Date:   Fri, 9 Jul 2021 17:33:36 +0200
From:   Geert Uytterhoeven <geert@...ux-m68k.org>
To:     Andrew Lunn <andrew@...n.ch>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Russell King <linux@...linux.org.uk>
Cc:     netdev <netdev@...r.kernel.org>,
        Linux-Renesas <linux-renesas-soc@...r.kernel.org>
Subject: PHY reset may still be asserted during MDIO probe

Hi all,

I'm investigating a network failure after kexec on the Renesas Koelsch
and Salvator-XS development boards, using the sh-eth or ravb driver.

During normal boot, the Ethernet interface is working fine:

    libphy: get_phy_c22_id:814: sh_mii: mdiobus_read() MII_PHYSID1 returned 34
    libphy: get_phy_c22_id:824: sh_mii: mdiobus_read() MII_PHYSID2 returned 5431
    libphy: get_phy_c22_id:832: sh_mii: phy_id = 0x00221537
    libphy: get_phy_device:895: sh_mii: get_phy_c22_id() returned 0
    fwnode_mdiobus_register_phy:109: sh_mii: get_phy_device() returned (ptrval)
    fwnode_mdiobus_phy_device_register:46: sh_mii: fwnode_irq_get() returned 191
    libphy: mdiobus_register_gpiod:48: mdiodev->reset_gpio = (ptrval)
    mdio_bus ee700000.ethernet-ffffffff:01:
mdiobus_register_device:88: assert MDIO reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 1)
    mdio_bus ee700000.ethernet-ffffffff:01: phy_device_register:931:
deassert PHY reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 0)
    Micrel KSZ8041RNLI ee700000.ethernet-ffffffff:01: phy_probe:3026:
deassert PHY reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 0)
    fwnode_mdiobus_phy_device_register:75: sh_mii:
phy_device_register() returned 0
    fwnode_mdiobus_register_phy:137: sh_mii:
fwnode_mdiobus_phy_device_register() returned 0
    of_mdiobus_register:188: of_mdiobus_register_phy(sh_mii,
/soc/ethernet@...00000/ethernet-phy@1, 1) returned 0
    sh-eth ee700000.ethernet eth0: Base address at 0xee700000,
2e:09:0a:00:6d:85, IRQ 126.

When using kexec, the PHY reset is asserted before starting the
new kernel:

    Micrel KSZ8041RNLI ee700000.ethernet-ffffffff:01: phy_detach:1759:
assert PHY reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 1)
    kexec_core: Starting new kernel
    Bye!

The new kernel fails to probe the PHY, as the PHY reset is still
asserted:

    libphy: get_phy_c22_id:814: sh_mii: mdiobus_read() MII_PHYSID1
returned 65535
    libphy: get_phy_c22_id:824: sh_mii: mdiobus_read() MII_PHYSID2
returned 65535
    libphy: get_phy_c22_id:832: sh_mii: phy_id = 0xffffffff
    libphy: get_phy_device:895: sh_mii: get_phy_c22_id() returned -19
    fwnode_mdiobus_register_phy:109: sh_mii: get_phy_device() returned -ENODEV
    of_mdiobus_register:188: of_mdiobus_register_phy(sh_mii,
/soc/ethernet@...00000/ethernet-phy@1, 1) returned -19
    mdio_bus ee700000.ethernet-ffffffff: MDIO device at address 1 is missing.
    sh-eth ee700000.ethernet eth0: Base address at 0xee700000,
2e:09:0a:00:6d:85, IRQ 126.

This issue can also be reproduced using unbind:

    # echo ee700000.ethernet > /sys/bus/platform/drivers/sh-eth/unbind
    sh-eth ee700000.ethernet eth0: Link is Down
    Micrel KSZ8041RNLI ee700000.ethernet-ffffffff:01: phy_detach:1759:
assert PHY reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 1)
    Micrel KSZ8041RNLI ee700000.ethernet-ffffffff:01: phy_remove:3120:
assert PHY reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 1)
    mdio_bus ee700000.ethernet-ffffffff:01: phy_device_remove:974:
assert PHY reset
    libphy: mdio_device_reset:124: calling gpiod_set_value_cansleep(..., 1)

and bind:

    # echo ee700000.ethernet > /sys/bus/platform/drivers/sh-eth/bind
    (same log as kexec boot)

I think fwnode_mdiobus_register_phy() should do the PHY reset (assert +
deassert) before calling get_phy_device(), but currently that happens
in phy_device_register(), which is called later.

Thanks for your comments!

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@...ux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds