linux-kernel - Re: [PATCH v3] PCI/ERR: Use pcie_aer_is_native() to judge whether OS owns AER

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]

Message-ID: <b54b068b-fe9a-8609-3e9f-170579affc27@bytedance.com>
Date:   Wed, 27 Jul 2022 17:37:23 +0800
From:   Zhuo Chen <chenzhuo.1@...edance.com>
To:     Sathyanarayanan Kuppuswamy 
        <sathyanarayanan.kuppuswamy@...ux.intel.com>
Cc:     ruscur@...sell.cc, oohall@...il.com, bhelgaas@...gle.com,
        lukas@...ner.de, jan.kiszka@...mens.com, stuart.w.hayes@...il.com,
        linuxppc-dev@...ts.ozlabs.org, linux-pci@...r.kernel.org,
        linux-kernel@...r.kernel.org, chenzhuo.1@...edance.com
Subject: Re: [PATCH v3] PCI/ERR: Use pcie_aer_is_native() to judge whether OS
 owns AER



On 7/26/22 1:35 PM, Zhuo Chen wrote:
> 
> On 7/26/22 9:02 PM, Sathyanarayanan Kuppuswamy wrote:
>>
>>
>> On 7/26/22 8:53 PM, Zhuo Chen wrote:
>>> Use pcie_aer_is_native() in place of "host->native_aer ||
>>> pcie_ports_native" to judge whether OS has native control of AER
>>> in pcie_do_recovery().
>>>
>>> Replace "dev->aer_cap && (pcie_ports_native || host->native_aer)" in
>>> get_port_device_capability() with pcie_aer_is_native(), which has no
>>> functional changes.
>>>
>>> Signed-off-by: Zhuo Chen <chenzhuo.1@...edance.com>
>>> ---
>>
>> Patch looks better now. It looks like following two changes
>> can also be replaced with pcie_aer_is_native() check.
>>
>> drivers/pci/pcie/aer.c:1407:    if ((host->native_aer || 
>> pcie_ports_native) && aer) {
>> drivers/pci/pcie/aer.c:1426:    if ((host->native_aer || 
>> pcie_ports_native) && aer) {
> 
> Good advice. But I wonder is there a scenario that dev->rcec ("root") is 
> NULL meanwhile dev->aer_cap is not NULL? If so, replace 
> "(host->native_aer || pcie_ports_native) && aer" with 
> pcie_aer_is_native() will change original function.
> 
Do you mean changing "if ((host->native_aer || pcie_ports_native) && 
aer)" into "if (pcie_aer_is_native(dev) && aer)" ?
I thought changing into "if (pcie_aer_is_native(dev))" before.

One another doubt. Not every pci device support aer. When dev->aer_cap 
is NULL and root->aer_cap is not NULL in aer_root_reset(), 
pcie_aer_is_native() will return false and OS cannot operate root 
register. It's different from just using "(host->native_aer || 
pcie_ports_native)".

Or we can change "if ((host->native_aer || pcie_ports_native) && aer)" 
into "if (pcie_aer_is_native(root))". But in this way, argument NULL 
pointer check should be added in pcie_aer_is_native().

>>
>>
>>
>>> Changelog:
>>> v3:
>>> - Simplify why we use pcie_aer_is_native().
>>> - Revert modification of pci_aer_clear_nonfatal_status() and comments.
>>> v2:
>>> - Add details and note in commit log.
>>> ---
>>>   drivers/pci/pcie/err.c          | 3 +--
>>>   drivers/pci/pcie/portdrv_core.c | 3 +--
>>>   2 files changed, 2 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
>>> index 0c5a143025af..121a53338e44 100644
>>> --- a/drivers/pci/pcie/err.c
>>> +++ b/drivers/pci/pcie/err.c
>>> @@ -184,7 +184,6 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev 
>>> *dev,
>>>       int type = pci_pcie_type(dev);
>>>       struct pci_dev *bridge;
>>>       pci_ers_result_t status = PCI_ERS_RESULT_CAN_RECOVER;
>>> -    struct pci_host_bridge *host = pci_find_host_bridge(dev->bus);
>>>       /*
>>>        * If the error was detected by a Root Port, Downstream Port, 
>>> RCEC,
>>> @@ -243,7 +242,7 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev 
>>> *dev,
>>>        * it is responsible for clearing this status.  In that case, the
>>>        * signaling device may not even be visible to the OS.
>>>        */
>>> -    if (host->native_aer || pcie_ports_native) {
>>> +    if (pcie_aer_is_native(dev)) {
>>>           pcie_clear_device_status(dev);
>>>           pci_aer_clear_nonfatal_status(dev);
>>>       }
>>> diff --git a/drivers/pci/pcie/portdrv_core.c 
>>> b/drivers/pci/pcie/portdrv_core.c
>>> index 604feeb84ee4..98c18f4a01b2 100644
>>> --- a/drivers/pci/pcie/portdrv_core.c
>>> +++ b/drivers/pci/pcie/portdrv_core.c
>>> @@ -221,8 +221,7 @@ static int get_port_device_capability(struct 
>>> pci_dev *dev)
>>>       }
>>>   #ifdef CONFIG_PCIEAER
>>> -    if (dev->aer_cap && pci_aer_available() &&
>>> -        (pcie_ports_native || host->native_aer)) {
>>> +    if (pcie_aer_is_native(dev) && pci_aer_available()) {
>>>           services |= PCIE_PORT_SERVICE_AER;
>>>           /*
>>
> 
> Thanks,
> Zhuo Chen

-- 
Thanks,
Zhuo Chen