Re: [PATCH] PCI/ERR: Fix run error recovery callbacks for all affected devices

[Dongdong Liu <liudongdong3@huawei.com>]

Hi Sinan
在 2019/1/26 1:09, Sinan Kaya 写道:
> On 1/25/2019 9:28 AM, Dongdong Liu wrote:
>> I want to fix 2 points by the patch.
>>
>> 1. For EP devices (such as multi-function EP device) under the same bus,
>> when one of the EP devices met non-fatal error, should report non-fatal
>> error only to the error endpoint device, no need to broadcast all of them.
>> That is the patch (PCI/AER: Report non-fatal errors only to the affected endpoint  #4.15)
>> have done, but current code PATCH [1] broken this.
>>
>> 2. We found a NULL pointer dereference issue for 74:02.0 device
>> when the device met a non-fatal error (firmware-first) after 4.19.
>> The topology is as below.
>
> Is it possible to split the patches into two to address these two different
> issues?

Good suggestion, will do.

>
> I can understand the first one but second one seems some mystery that needs
> to be explored in detail.
>

  +-[0000:74]-+-02.0  Huawei Technologies Co., Ltd. HiSilicon SAS 3.0 HBA
  |           \-03.0  Huawei Technologies Co., Ltd. HiSilicon AHCI HBA

74:02.0 is a RCiEP, but do not under a root port.
Current code have the issue as the below code, see [DD].
aer_recover_work_func()
     pcie_do_recovery() {

	/*
	 * Error recovery runs on all subordinates of the first downstream port.
	 * If the downstream port detected the error, it is cleared at the end.
	 */
	if (!(pci_pcie_type(dev) == PCI_EXP_TYPE_ROOT_PORT ||
	      pci_pcie_type(dev) == PCI_EXP_TYPE_DOWNSTREAM))
		dev = dev->bus->self;   //[DD]: Here dev will be NULL pointer for 74:02.0（RCiEP）as it does not have a root port.
	bus = dev->subordinate;
  }

We do not have the  NULL pointer dereference issue
before the patch [1] PCI/ERR: Run error recovery callbacks for all affected devices   #4.20 .

Thanks,
Dongdong
> .
>