Message-ID: <19d73292-95c4-2990-8ec6-c2a9b87c5598@applied-asynchrony.com>
Date:   Tue, 29 Jan 2019 17:57:38 +0100
From:   Holger Hoffstätte <holger@...lied-asynchrony.com>
To:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Camelia Alexandra Groza <camelia.groza@....com>
Cc:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "stable@...r.kernel.org" <stable@...r.kernel.org>,
        Scott Wood <oss@...error.net>,
        "David S. Miller" <davem@...emloft.net>,
        Heiner Kallweit <hkallweit1@...il.com>,
        Andrew Lunn <andrew@...n.ch>
Subject: Re: [PATCH 4.19 006/103] net: phy: phy driver features are mandatory

On 1/29/19 5:33 PM, Greg Kroah-Hartman wrote:
> On Tue, Jan 29, 2019 at 04:05:16PM +0000, Camelia Alexandra Groza wrote:
>>> -----Original Message-----
>>> From: Holger Hoffstätte <holger@...lied-asynchrony.com>
>>> Sent: Tuesday, January 29, 2019 17:50
>>> To: Greg Kroah-Hartman <gregkh@...uxfoundation.org>; linux-
>>> kernel@...r.kernel.org
>>> Cc: stable@...r.kernel.org; Scott Wood <oss@...error.net>; Camelia
>>> Alexandra Groza <camelia.groza@....com>; David S. Miller
>>> <davem@...emloft.net>; Heiner Kallweit <hkallweit1@...il.com>
>>> Subject: Re: [PATCH 4.19 006/103] net: phy: phy driver features are
>>> mandatory
>>>
>>> On 1/29/19 12:34 PM, Greg Kroah-Hartman wrote:
>>>> 4.19-stable review patch.  If anyone has any objections, please let me
>>> know.
>>>>
>>>> ------------------
>>>>
>>>> From: Camelia Groza <camelia.groza@....com>
>>>>
>>>> [ Upstream commit 3e64cf7a435ed0500e3adaa8aada2272d3ae8abc ]
>>>>
>>>> Since phy driver features became a link_mode bitmap, phy drivers that
>>>> don't have a list of features configured will cause the kernel to crash
>>>> when probed.
>>>>
>>>> Prevent the phy driver from registering if the features field is missing.
>>>>
>>>> Fixes: 719655a14971 ("net: phy: Replace phy driver features u32 with
>>> link_mode bitmap")
>>>> Reported-by: Scott Wood <oss@...error.net>
>>>> Signed-off-by: Camelia Groza <camelia.groza@....com>
>>>> Signed-off-by: David S. Miller <davem@...emloft.net>
>>>> Signed-off-by: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
>>>> ---
>>>>    drivers/net/phy/phy_device.c |    5 +++++
>>>>    include/linux/phy.h          |    4 ++--
>>>>    2 files changed, 7 insertions(+), 2 deletions(-)
>>>>
>>>> --- a/drivers/net/phy/phy_device.c
>>>> +++ b/drivers/net/phy/phy_device.c
>>>> @@ -1917,6 +1917,11 @@ int phy_driver_register(struct phy_drive
>>>>    {
>>>>    	int retval;
>>>>
>>>> +	if (WARN_ON(!new_driver->features)) {
>>>> +		pr_err("%s: Driver features are missing\n", new_driver-
>>>> name);
>>>> +		return -EINVAL;
>>>> +	}
>>>> +
>>>>    	new_driver->mdiodrv.flags |= MDIO_DEVICE_IS_PHY;
>>>>    	new_driver->mdiodrv.driver.name = new_driver->name;
>>>>    	new_driver->mdiodrv.driver.bus = &mdio_bus_type;
>>>> --- a/include/linux/phy.h
>>>> +++ b/include/linux/phy.h
>>>> @@ -491,8 +491,8 @@ struct phy_device {
>>>>     *   only works for PHYs with IDs which match this field
>>>>     * name: The friendly name of this PHY type
>>>>     * phy_id_mask: Defines the important bits of the phy_id
>>>> - * features: A list of features (speed, duplex, etc) supported
>>>> - *   by this PHY
>>>> + * features: A mandatory list of features (speed, duplex, etc)
>>>> + *   supported by this PHY
>>>>     * flags: A bitfield defining certain other features this PHY
>>>>     *   supports (like interrupts)
>>>>     *
>>>>
>>>
>>> This patch leads to the following dump during boot and no working
>>> network:
>>>
>>> Jan 29 13:11:50 ragnarok kernel: RIP: 0010:phy_driver_register+0x5f/0x80
>>> [libphy]
>>> Jan 29 13:11:50 ragnarok kernel: Code: dd 06 a0 48 c7 47 40 30 e4 06 a0 48 89
>>> 77 10 c7 47 24 02 00 00 00 e8 10 5c 44 e1 89 c5 85 c0 0f 85 aa 1d 00 00 89 e8 5b
>>> 5d c3 <0f> 0b 48 c7 c7 f0 20 07 a0 48 89 c6 bd ea ff ff ff e8 c9 4c 02 e1
>>> Jan 29 13:11:50 ragnarok kernel: RSP: 0018:ffffc90000df3c78 EFLAGS:
>>> 00010246
>>> Jan 29 13:11:50 ragnarok kernel: RAX: ffffffffa0071c8d RBX: 0000000000000000
>>> RCX: 0000000000000000
>>> Jan 29 13:11:50 ragnarok kernel: RDX: 0000000000000000 RSI: ffffffffa0076900
>>> RDI: ffffffffa0076040
>>> Jan 29 13:11:50 ragnarok kernel: RBP: ffffffffa007a000 R08: 0000000000000044
>>> R09: 0000000000000228
>>> Jan 29 13:11:50 ragnarok kernel: R10: ffff888603a0d6f8 R11:
>>> 000000000000003c R12: ffff888601252d00
>>> Jan 29 13:11:50 ragnarok kernel: R13: 0000000000000002 R14:
>>> ffffc90000df3e98 R15: ffffffffa0076900
>>> Jan 29 13:11:50 ragnarok kernel: FS:  00007feea9ced800(0000)
>>> GS:ffff888607240000(0000) knlGS:0000000000000000
>>> Jan 29 13:11:50 ragnarok kernel: CS:  0010 DS: 0000 ES: 0000 CR0:
>>> 0000000080050033
>>> Jan 29 13:11:50 ragnarok kernel: CR2: 00007feea9cdfc42 CR3:
>>> 00000005fec64001 CR4: 00000000000606e0
>>> Jan 29 13:11:50 ragnarok kernel: Call Trace:
>>> Jan 29 13:11:50 ragnarok kernel:  ? 0xffffffffa007a000
>>> Jan 29 13:11:50 ragnarok kernel:  phy_init+0x24/0x58 [libphy]
>>> Jan 29 13:11:50 ragnarok kernel:  do_one_initcall+0x43/0x1af
>>> Jan 29 13:11:50 ragnarok kernel:  ? _cond_resched+0x15/0x30
>>> Jan 29 13:11:50 ragnarok kernel:  ? kmem_cache_alloc_trace+0x154/0x1b0
>>> Jan 29 13:11:50 ragnarok kernel:  do_init_module+0x5a/0x210
>>> Jan 29 13:11:50 ragnarok kernel:  load_module+0x203a/0x2500
>>> Jan 29 13:11:50 ragnarok kernel:  ? vfs_read+0x10d/0x130
>>> Jan 29 13:11:50 ragnarok kernel:  ? __se_sys_finit_module+0xb1/0xd0
>>> Jan 29 13:11:50 ragnarok kernel:  __se_sys_finit_module+0xb1/0xd0
>>> Jan 29 13:11:50 ragnarok kernel:  do_syscall_64+0x3e/0xe0
>>> Jan 29 13:11:50 ragnarok kernel:
>>> entry_SYSCALL_64_after_hwframe+0x44/0xa9
>>> Jan 29 13:11:50 ragnarok kernel: RIP: 0033:0x7feeaa2417c9
>>> Jan 29 13:11:50 ragnarok kernel: Code: 00 00 00 75 05 48 83 c4 18 c3 e8 82 9c 01
>>> 00 66 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05
>>> <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 87 c6 0c 00 f7 d8 64 89 01 48
>>> Jan 29 13:11:50 ragnarok kernel: RSP: 002b:00007fff6fa6ea08 EFLAGS:
>>> 00000246 ORIG_RAX: 0000000000000139
>>> Jan 29 13:11:50 ragnarok kernel: RAX: ffffffffffffffda RBX: 000055c6eac06db0
>>> RCX: 00007feeaa2417c9
>>> Jan 29 13:11:50 ragnarok kernel: RDX: 0000000000000000 RSI:
>>> 00007feeaa4198fd RDI: 0000000000000006
>>> Jan 29 13:11:50 ragnarok kernel: RBP: 00007feeaa4198fd R08:
>>> 0000000000000000 R09: 000055c6eac03000
>>> Jan 29 13:11:50 ragnarok kernel: R10: 0000000000000006 R11:
>>> 0000000000000246 R12: 0000000000000000
>>> Jan 29 13:11:50 ragnarok kernel: R13: 000055c6eabfc1d0 R14:
>>> 0000000000020000 R15: 000055c6eac06db0
>>> Jan 29 13:11:50 ragnarok kernel: ---[ end trace 18e05a0afeb140b2 ]---
>>> Jan 29 13:11:50 ragnarok kernel: libphy: Generic 10G PHY: Driver features are
>>> missing
>>>
>>> Apparently the r8169 driver for my NIC does something wrong wrt. this
>>> fix and the original commit (719655a14971), but regardless - reverting this
>>> patch leads to a booting & working 4.19.19:
>>>
>>> Jan 29 13:29:16 ragnarok kernel: libphy: r8169: probed
>>> Jan 29 13:29:19 ragnarok kernel: RTL8211E Gigabit Ethernet r8169-400:00:
>>> attached PHY driver [RTL8211E Gigabit Ethernet] (mii_bus:phy_addr=r8169-
>>> 400:00, irq=IGNORE)
>>>
>>> I've cc'ed Heiner Kallweit who might know why phylib with this fix and r8169
>>> don't like each other.
>>>
>>> cheers,
>>> Holger
>>
>> Hi
>>
>> This points to an issue with the Generic 10G PHY driver. The following patch seems to be missing from the 4.19 stable tree and should be merged:
>> f802912 net: phy: genphy_10g_driver: Avoid NULL pointer dereference
> 
> But that patch says it:
> Fixes: 719655a14971 ("net: phy: Replace phy driver features u32 with link_mode bitmap")
> 
> Which is not in 4.19.y, it showed up in 4.20.

Nope, as you found 719655a14971 was backported:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=linux-4.19.y&id=719655a149715f26fc4de904fe0aa83068bd5b9e

> Ah, wait, I think that there was a patch backported to the 4.19.y queue
> that should not have, specifically:
> 	3e64cf7a435e ("net: phy: phy driver features are mandatory")
> which said it too fixes:
> 	719655a14971 ("net: phy: Replace phy driver features u32 with link_mode bitmap")
> 
> 
> Holger, if you revert 3e64cf7a435e, does that solve the problem for you
> here?

Yes, as per $SUBJECT if I patch --reverse 3e64cf7a435e it works again,
seem my original log above.

Fwiw just test-booted 5.0-rc4 and it works too.

cheers,
Holger

Powered by blists - more mailing lists