MIME-Version: 1.0
In-Reply-To: <502375193.icN6k8xHrH@tjmaciei-mobl1>
References: <20170829182032.61961-1-ppenkov@google.com> <502375193.icN6k8xHrH@tjmaciei-mobl1>
From: Eric Dumazet <edumazet@google.com>
Date: Sat, 30 Sep 2017 13:51:17 -0700
Message-ID: <CANn89iJ=fSR8OywcN-WomHta=3QW01CMLBJzSev53Vm=wtsJxQ@mail.gmail.com>
Subject: Re: [v1] iov_iter: fix page_copy_sane for compound pages
To: Thiago Macieira <thiago.macieira@intel.com>
Cc: LKML <linux-kernel@vger.kernel.org>, Al Viro <viro@zeniv.linux.org.uk>,
        Petar Penkov <peterpenkov96@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1230
Lines: 31

On Sat, Sep 30, 2017 at 12:24 PM, Thiago Macieira
<thiago.macieira@intel.com> wrote:
> On Tuesday, 29 August 2017 11:20:32 PDT Petar Penkov wrote:
>> Issue is that if the data crosses a page boundary inside a compound
>> page, this check will incorrectly trigger a WARN_ON.
>>
>> To fix this, compute the order using the head of the compound page and
>> adjust the offset to be relative to that head.
>>
>> Fixes: 72e809ed81ed ("iov_iter: sanity checks for copy to/from page
>> primitives")
>
> Hello
>
> Is this patch slated to end up in one of the 4.13.x updates? It landed on
> v4.14-rc2 already but seems to have missed the 4.13.3 and 4.13.4 tagging.
>
> Without this patch, I can't connect any USB Ethernet or the kernel will start
> producing that WARN_ON message and returning -EFAULT for quite a few programs.
>

Hmm... problem Petar had originally has been solved.

His commit ( 90e33d45940793def6f773b2d528e9f3c84ffdc7 in Dave Miller net-next)
no longer hits the problem, while his prior version triggered the issue.

Relevant and updated part is :

+               page = virt_to_head_page(data);
+               offset = data - page_address(page);
+               skb_fill_page_desc(skb, i - 1, page, offset, fragsz);