Linux RDMA and InfiniBand development
 help / color / mirror / Atom feed
From: Aleksandr Mishin <amishin@t-argos.ru>
To: Junxian Huang <huangjunxian6@hisilicon.com>,
	Leon Romanovsky <leon@kernel.org>
Cc: Wei Xu <xuwei5@hisilicon.com>,
	Chengchang Tang <tangchengchang@huawei.com>,
	Jason Gunthorpe <jgg@ziepe.ca>, Xi Wang <wangxi11@huawei.com>,
	Shengming Shu <shushengming1@huawei.com>,
	Weihang Li <liweihang@huawei.com>, <linux-rdma@vger.kernel.org>,
	<linux-kernel@vger.kernel.org>, <lvc-project@linuxtesting.org>
Subject: Re: [PATCH] RDMA: hns: Fix possible null pointer dereference
Date: Tue, 9 Apr 2024 21:01:38 +0300	[thread overview]
Message-ID: <1a96d3be-cc40-4aef-bd4d-bb3a9ef3a7d4@t-argos.ru> (raw)
In-Reply-To: <1476bb96-c0a9-379a-546b-98fc8a06beea@hisilicon.com>


Thank you! I assumed something like this, but I couldn't find any 
confirmations and offered a patch as a solution.

On 09.04.2024 14:10, Junxian Huang wrote:
> 
> 
> On 2024/4/9 17:26, Leon Romanovsky wrote:
>> On Tue, Apr 09, 2024 at 11:30:47AM +0300, Aleksandr Mishin wrote:
>>> In hns_roce_hw_v2_get_cfg() pci_match_id() may return
>>> NULL which is later dereferenced. Fix this bug by adding NULL check.
>>
>> I don't know, this NULL can't happen in this flow.
>>
>> Thanks
>>
> 
> Yeah, it's already checked here:
> 
> 6911 static int hns_roce_hw_v2_init_instance(struct hnae3_handle *handle)
> 6912 {
> 6913         const struct hnae3_ae_ops *ops = handle->ae_algo->ops;
> 6914         const struct pci_device_id *id;
> 6915         struct device *dev = &handle->pdev->dev;
> 6916         int ret;
> 6917
> 6918         handle->rinfo.instance_state = HNS_ROCE_STATE_INIT;
> 6919
> 6920         if (ops->ae_dev_resetting(handle) || ops->get_hw_reset_stat(handle)) {
> 6921                 handle->rinfo.instance_state = HNS_ROCE_STATE_NON_INIT;
> 6922                 goto reset_chk_err;
> 6923         }
> 6924
> 6925         id = pci_match_id(hns_roce_hw_v2_pci_tbl, handle->pdev);
> 6926         if (!id)
> 6927                 return 0;
> 6928
> 6929         if (id->driver_data && handle->pdev->revision == PCI_REVISION_ID_HIP08)
> 6930                 return 0;
> 6931
> 6932         ret = __hns_roce_hw_v2_init_instance(handle);
> 
> Junxian
> 
>>>
>>> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>>>
>>> Fixes: 0b567cde9d7a ("RDMA/hns: Enable RoCE on virtual functions")
>>> Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru>
>>> ---
>>>   drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 13 +++++++++++--
>>>   1 file changed, 11 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
>>> index ba7ae792d279..31a2093334d9 100644
>>> --- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
>>> +++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
>>> @@ -6754,7 +6754,7 @@ static const struct pci_device_id hns_roce_hw_v2_pci_tbl[] = {
>>>   
>>>   MODULE_DEVICE_TABLE(pci, hns_roce_hw_v2_pci_tbl);
>>>   
>>> -static void hns_roce_hw_v2_get_cfg(struct hns_roce_dev *hr_dev,
>>> +static int hns_roce_hw_v2_get_cfg(struct hns_roce_dev *hr_dev,
>>>   				  struct hnae3_handle *handle)
>>>   {
>>>   	struct hns_roce_v2_priv *priv = hr_dev->priv;
>>> @@ -6763,6 +6763,9 @@ static void hns_roce_hw_v2_get_cfg(struct hns_roce_dev *hr_dev,
>>>   
>>>   	hr_dev->pci_dev = handle->pdev;
>>>   	id = pci_match_id(hns_roce_hw_v2_pci_tbl, hr_dev->pci_dev);
>>> +	if (!id)
>>> +		return -ENXIO;
>>> +
>>>   	hr_dev->is_vf = id->driver_data;
>>>   	hr_dev->dev = &handle->pdev->dev;
>>>   	hr_dev->hw = &hns_roce_hw_v2;
>>> @@ -6789,6 +6792,8 @@ static void hns_roce_hw_v2_get_cfg(struct hns_roce_dev *hr_dev,
>>>   
>>>   	hr_dev->reset_cnt = handle->ae_algo->ops->ae_dev_reset_cnt(handle);
>>>   	priv->handle = handle;
>>> +
>>> +	return 0;
>>>   }
>>>   
>>>   static int __hns_roce_hw_v2_init_instance(struct hnae3_handle *handle)
>>> @@ -6806,7 +6811,11 @@ static int __hns_roce_hw_v2_init_instance(struct hnae3_handle *handle)
>>>   		goto error_failed_kzalloc;
>>>   	}
>>>   
>>> -	hns_roce_hw_v2_get_cfg(hr_dev, handle);
>>> +	ret = hns_roce_hw_v2_get_cfg(hr_dev, handle);
>>> +	if (ret) {
>>> +		dev_err(hr_dev->dev, "RoCE Engine cfg failed!\n");
>>> +		goto error_failed_roce_init;
>>> +	}
>>>   
>>>   	ret = hns_roce_init(hr_dev);
>>>   	if (ret) {
>>> -- 
>>> 2.30.2
>>>
> 

-- 
Kind regards
Aleksandr

      reply	other threads:[~2024-04-09 18:07 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-04-09  8:30 [PATCH] RDMA: hns: Fix possible null pointer dereference Aleksandr Mishin
2024-04-09  9:26 ` Leon Romanovsky
2024-04-09 11:10   ` Junxian Huang
2024-04-09 18:01     ` Aleksandr Mishin [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1a96d3be-cc40-4aef-bd4d-bb3a9ef3a7d4@t-argos.ru \
    --to=amishin@t-argos.ru \
    --cc=huangjunxian6@hisilicon.com \
    --cc=jgg@ziepe.ca \
    --cc=leon@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-rdma@vger.kernel.org \
    --cc=liweihang@huawei.com \
    --cc=lvc-project@linuxtesting.org \
    --cc=shushengming1@huawei.com \
    --cc=tangchengchang@huawei.com \
    --cc=wangxi11@huawei.com \
    --cc=xuwei5@hisilicon.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox