Subject: Re: [PATCH] PCI: endpoint: Fix use after free in pci_epf_remove_cfs()
To:     Shunsuke Mie <mie@igel.co.jp>
CC:     <lorenzo.pieralisi@arm.com>, <bhelgaas@google.com>,
        <linux-pci@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
        <stable@vger.kernel.org>
References: <20210621070058.37682-1-mie@igel.co.jp>
From:   Kishon Vijay Abraham I <kishon@ti.com>
Message-ID: <ed4de030-e0b3-503f-f0dc-75b103769dfc@ti.com>
Date:   Tue, 7 Dec 2021 12:56:41 +0530
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.11.0
MIME-Version: 1.0
In-Reply-To: <20210621070058.37682-1-mie@igel.co.jp>
Content-Type: text/plain; charset="utf-8"
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

Hi Shunsuke,

On 21/06/21 12:30 pm, Shunsuke Mie wrote:
> All of entries are freed in a loop, however, the freed entry is accessed
> by list_del() after the loop.
> 
> When epf driver that includes pci-epf-test unload, the pci_epf_remove_cfs()
> is called, and occurred the use after free. Therefore, kernel panics
> randomly after or while the module unloading.
> 
> I tested this patch with r8a77951-Salvator-xs boards.

Can you provide the crash dump?
> 
> Fixes: ef1433f ("PCI: endpoint: Create configfs entry for each pci_epf_device_id table entry")
> Signed-off-by: Shunsuke Mie <mie@igel.co.jp>
> ---
>  drivers/pci/endpoint/pci-epf-core.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/endpoint/pci-epf-core.c b/drivers/pci/endpoint/pci-epf-core.c
> index e9289d10f822..538e902b0ba6 100644
> --- a/drivers/pci/endpoint/pci-epf-core.c
> +++ b/drivers/pci/endpoint/pci-epf-core.c
> @@ -202,8 +202,10 @@ static void pci_epf_remove_cfs(struct pci_epf_driver *driver)
>  		return;
>  
>  	mutex_lock(&pci_epf_mutex);
> -	list_for_each_entry_safe(group, tmp, &driver->epf_group, group_entry)
> +	list_for_each_entry_safe(group, tmp, &driver->epf_group, group_entry) {
> +		list_del(&group->group_entry);

Need full crash dump to see if this is really required or not. Ideally this
should be handled by configfs (or a configfs API could be used).

Thanks,
Kishon