netdev - Re: Re: [PATCH linux next] net:dsa:fix the dsa

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] [day] [month] [year] [list]

Message-Id: <20241224080532.3958498-1-peilinhe2020@163.com>
Date: Tue, 24 Dec 2024 08:05:32 +0000
From: PeilinHe <peilinhe2020@....com>
To: olteanv@...il.com
Cc: andrew@...n.ch,
	davem@...emloft.net,
	edumazet@...gle.com,
	fan.yu9@....com.cn,
	he.peilin@....com.cn,
	jiang.kun2@....com.cn,
	linux-kernel@...r.kernel.org,
	netdev@...r.kernel.org,
	qiu.yutan@....com.cn,
	tu.qiang35@....com.cn,
	wang.yaxin@....com.cn,
	xu.xin16@....com.cn,
	yang.yang29@....com.cn,
	ye.xingchen@....com.cn,
	zhang.yunkai@....com.cn
Subject: Re: Re: [PATCH linux next] net:dsa:fix the dsa_ptr null pointer dereference

>Thank you for the patch.
>
>There are many process problems with it however.
>
>The most glaring one is that you are examining a crash from kernel 5.4
>but patching linux-next, without having apparently also tested linux-next.
>It appears that you just made a static analysis which may result in
>incorrect conclusions. When submitting patches upstream you always have
>to test on the latest version and understand afterwards what is missing
>and needs to be backported in the particular stable version you are using.
>
>In particular here, dsa_switch_shutdown() now has this:
>
>	dsa_switch_for_each_user_port(dp, ds) {
>		conduit = dsa_port_to_conduit(dp);
>		user_dev = dp->user;
>
>		netif_device_detach(user_dev);
>		~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>		netdev_upper_dev_unlink(conduit, user_dev);
>	}
>
>After netif_device_detach() is called, my expectation is that
>ethnl_ops_begin() sees that netif_device_present() is false, so it
>returns -ENODEV and does not proceed further to call into the device's
>ethtool ops. So that eliminates the premise for the crash.
>
>Secondly, linux-next is not a kernel tree that accepts patches, it is
>just for integration. For netdev, we have net.git for bug fixes and
>net-next.git for new features. You have to target your patch to net.git
>by using "[PATCH net v1]".
>
>If the problem does not exist in net.git but exists in stable kernels,
>you have to identify which patches are missing, adapt them if necessary,
>and then send them to stable@...r.kernel.org, with netdev and the other
>maintainers also CCed, and with a subject prefix along the lines of
>"[PATCH stable 5.4]". Generally, backporting patches manually to stable
>is rarely needed, so if that needs to happen, please use the space under
>the "---" marker (this is discarded when applying the patch in git) to
>explain to maintainers why (what conflicted, if it simply appears to
>have been missed, etc).
>
>There are other things to be aware of in Documentation/process/, I just
>summarized to you what I considered most relevant here.

Thank you for your feedback.
First, I apologize for an error in my previous commit description. 
The kernel version that experienced the panic was actually 5.15.
Second, netif_device_present() does indeed prevent the null pointer 
dereference of dsa_ptr.
Finally, this issue still persists in the 5.15 stable release, 
therefore, following your suggestion, I will submit the relevant 
patch to the 5.15 stable branch.