Re: [RFT PATCH] xhci: Fix memory leak when caching protocol extended capability PSI tables

[Mathias Nyman]

On 11.2.2020 14.29, Mathias Nyman wrote:
> On 11.2.2020 14.23, Greg KH wrote:
>> On Tue, Feb 11, 2020 at 11:56:12AM +0100, Marek Szyprowski wrote:
>>> Hi
>>>
>>> On 08.01.2020 16:17, Mathias Nyman wrote:
>>>> xhci driver assumed that xHC controllers have at most one custom
>>>> supported speed table (PSI) for all usb 3.x ports.
>>>> Memory was allocated for one PSI table under the xhci hub structure.
>>>>
>>>> Turns out this is not the case, some controllers have a separate
>>>> "supported protocol capability" entry with a PSI table for each port.
>>>> This means each usb3 port can in theory support different custom speeds.
>>>>
>>>> To solve this cache all supported protocol capabilities with their PSI
>>>> tables in an array, and add pointers to the xhci port structure so that
>>>> every port points to its capability entry in the array.
>>>>
>>>> When creating the SuperSpeedPlus USB Device Capability BOS descriptor
>>>> for the xhci USB 3.1 roothub we for now will use only data from the
>>>> first USB 3.1 capable protocol capability entry in the array.
>>>> This could be improved later, this patch focuses resolving
>>>> the memory leak.
>>>>
>>>> Reported-by: Paul Menzel <pmenzel@xxxxxxxxxxxxx>
>>>> Reported-by: Sajja Venkateswara Rao <VenkateswaraRao.Sajja@xxxxxxx>
>>>> Signed-off-by: Mathias Nyman <mathias.nyman@xxxxxxxxxxxxxxx>
>>>
>>> This patch landed in today's linux-next (20200211) and causes NULL 
>>> pointer dereference during second suspend/resume cycle on Samsung 
>>> Exynos5422-based (arm 32bit) Odroid XU3lite board:
>>>
>>> # time rtcwake -s10 -mmem
>>> rtcwake: wakeup from "mem" using /dev/rtc0 at Tue Feb 11 10:51:43 2020
>>> PM: suspend entry (deep)
>>> Filesystems sync: 0.012 seconds
>>> Freezing user space processes ... (elapsed 0.010 seconds) done.
>>> OOM killer disabled.
>>> Freezing remaining freezable tasks ... (elapsed 0.002 seconds) done.
>>> smsc95xx 1-1.1:1.0 eth0: entering SUSPEND2 mode
>>> wake enabled for irq 153
>>> wake enabled for irq 158
>>> samsung-pinctrl 13400000.pinctrl: Setting external wakeup interrupt 
>>> mask: 0xffffffe7
>>> Disabling non-boot CPUs ...
>>> IRQ 51: no longer affine to CPU1
>>> IRQ 52: no longer affine to CPU2
>>> s3c2410-wdt 101d0000.watchdog: watchdog disabled
>>> wake disabled for irq 158
>>> usb usb1: root hub lost power or was reset
>>> usb usb2: root hub lost power or was reset
>>> wake disabled for irq 153
>>> exynos-tmu 10060000.tmu: More trip points than supported by this TMU.
>>> exynos-tmu 10060000.tmu: 2 trip points should be configured in polling mode.
>>> exynos-tmu 10064000.tmu: More trip points than supported by this TMU.
>>> exynos-tmu 10064000.tmu: 2 trip points should be configured in polling mode.
>>> exynos-tmu 10068000.tmu: More trip points than supported by this TMU.
>>> exynos-tmu 10068000.tmu: 2 trip points should be configured in polling mode.
>>> exynos-tmu 1006c000.tmu: More trip points than supported by this TMU.
>>> exynos-tmu 1006c000.tmu: 2 trip points should be configured in polling mode.
>>> exynos-tmu 100a0000.tmu: More trip points than supported by this TMU.
>>> exynos-tmu 100a0000.tmu: 6 trip points should be configured in polling mode.
>>> usb usb3: root hub lost power or was reset
>>> s3c-rtc 101e0000.rtc: rtc disabled, re-enabling
>>> usb usb4: root hub lost power or was reset
>>> xhci-hcd xhci-hcd.8.auto: No ports on the roothubs?
>>> PM: dpm_run_callback(): platform_pm_resume+0x0/0x44 returns -12
>>> PM: Device xhci-hcd.8.auto failed to resume async: error -12
>>> hub 3-0:1.0: hub_ext_port_status failed (err = -32)
>>> hub 4-0:1.0: hub_ext_port_status failed (err = -32)
>>> usb 1-1: reset high-speed USB device number 2 using exynos-ehci
>>> usb 1-1.1: reset high-speed USB device number 3 using exynos-ehci
>>> OOM killer enabled.
>>> Restarting tasks ... done.
>>>
>>> realÂÂÂ 0m11.890s
>>> userÂÂÂ 0m0.001s
>>> sysÂÂÂÂ 0m0.679s
>>> root@target:~# PM: suspend exit
>>> mmc_host mmc0: Bus speed (slot 0) = 50000000Hz (slot req 400000Hz, 
>>> actual 396825HZ div = 63)
>>> mmc_host mmc0: Bus speed (slot 0) = 200000000Hz (slot req 200000000Hz, 
>>> actual 200000000HZ div = 0)
>>> mmc_host mmc0: Bus speed (slot 0) = 50000000Hz (slot req 52000000Hz, 
>>> actual 50000000HZ div = 0)
>>> mmc_host mmc0: Bus speed (slot 0) = 400000000Hz (slot req 200000000Hz, 
>>> actual 200000000HZ div = 1)
>>> smsc95xx 1-1.1:1.0 eth0: link up, 100Mbps, full-duplex, lpa 0xC1E1
>>>
>>> root@target:~#
>>> root@target:~# time rtcwake -s10 -mmem[ÂÂ 35.451572] vdd_ldo12: disabling
>>>
>>> rtcwake: wakeup from "mem" using /dev/rtc0 at Tue Feb 11 10:52:02 2020
>>> PM: suspend entry (deep)
>>> Filesystems sync: 0.004 seconds
>>> Freezing user space processes ... (elapsed 0.006 seconds) done.
>>> OOM killer disabled.
>>> Freezing remaining freezable tasks ... (elapsed 0.070 seconds) done.
>>> hub 4-0:1.0: hub_ext_port_status failed (err = -32)
>>> hub 3-0:1.0: hub_ext_port_status failed (err = -32)
>>> 8<--- cut here ---
>>> Unable to handle kernel NULL pointer dereference at virtual address 00000014
>>> pgd = 4c26b54b
>>> [00000014] *pgd=00000000
>>> Internal error: Oops: 17 [#1] PREEMPT SMP ARM
>>> Modules linked in:
>>> CPU: 3 PID: 1468 Comm: kworker/u16:23 Not tainted 
>>> 5.6.0-rc1-next-20200211 #268
>>> Hardware name: Samsung Exynos (Flattened Device Tree)
>>> Workqueue: events_unbound async_run_entry_fn
>>> PC is at xhci_suspend+0x12c/0x520
>>> LR is at 0xa6aa9898
>>> pc : [<c0724c90>]ÂÂÂ lr : [<a6aa9898>]ÂÂÂ psr: 60000093
>>> sp : ec401df8 ip : 0000001a fp : c12e7864
>>> r10: 00000000Â r9 : ecfb87b0Â r8 : ecfb8220
>>> r7 : 00000000Â r6 : 00000000Â r5 : 00000004Â r4 : ecfb81f0
>>> r3 : 00007d00Â r2 : 00000001Â r1 : 00000001Â r0 : 00000000
>>> Flags: nZCv IRQs off FIQs on Mode SVC_32 ISA ARM Segment none
>>> Control: 10c5387d Table: 6bd4006a DAC: 00000051
>>> Process kworker/u16:23 (pid: 1468, stack limit = 0x6e4b6fba)
>>> Stack: (0xec401df8 to 0xec402000)
>>> ...
>>> [<c0724c90>] (xhci_suspend) from [<c061b4f4>] (dpm_run_callback+0xb4/0x3fc)
>>> [<c061b4f4>] (dpm_run_callback) from [<c061bd5c>] 
>>> (__device_suspend+0x134/0x7e8)
>>> [<c061bd5c>] (__device_suspend) from [<c061c42c>] (async_suspend+0x1c/0x94)
>>> [<c061c42c>] (async_suspend) from [<c0154bd0>] 
>>> (async_run_entry_fn+0x48/0x1b8)
>>> [<c0154bd0>] (async_run_entry_fn) from [<c0149b38>] 
>>> (process_one_work+0x230/0x7bc)
>>> [<c0149b38>] (process_one_work) from [<c014a108>] (worker_thread+0x44/0x524)
>>> [<c014a108>] (worker_thread) from [<c01511fc>] (kthread+0x130/0x164)
>>> [<c01511fc>] (kthread) from [<c01010b4>] (ret_from_fork+0x14/0x20)
>>> Exception stack(0xec401fb0 to 0xec401ff8)
>>> ...
>>> ---[ end trace c72caf6487666442 ]---
>>> note: kworker/u16:23[1468] exited with preempt_count 1
>>>
>>> Reverting it fixes the NULL pointer issue. I can provide more 
>>> information or do some other tests. Just let me know what will help to 
>>> fix it.
>>>
>>>  > ...
>>
>> Ugh.  Mathias, should I just revert this for now?
>>
> 
> Yes, revert it.
> 
> This looks very odd, after second resume, and losing power driver
> can't find any port at all.
> 
> Marek, do you still get the "xhci-hcd xhci-hcd.8.auto: No ports on the roothubs?"
> message on second resume after reverting the patch?
> 

Ok, I think I got it. 
Patch doesn't set xhci->num_port_caps to 0 in xhci_mem_cleanup().

Adding new ports will fail when we reinitialize xhci manually, like in this
exynos case where xhci loses power in suspend/resume cycle.  

I'll post a new version soon

-Mathias