| 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
| ||
|
Message-ID: <4CBC12EC.60109@bfs.de>
Date: Mon, 18 Oct 2010 11:27:08 +0200
From: walter harms <wharms@....de>
To: Julia Lawall <julia@...u.dk>
CC: kernel-janitors@...r.kernel.org, devel@...verdev.osuosl.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/3] drivers/staging/cx25821: Use kasprintf
Julia Lawall schrieb:
> On Mon, 18 Oct 2010, walter harms wrote:
>
>>
>> Julia Lawall schrieb:
>>> Rewrite the initialization of a dev field. In the original code, in each
>>> case there was a kmalloc followed by a memcpy, as illustrated by the
>>> semantic patch below. In the case that the provided string was the empty
>>> string, the allocated memory was then overwritten with a constant string,
>>> causing a memory leak. Finally, there was no provision for returning
>>> -ENOMEM in case of failure of the memory allocation. Indeed, the return
>>> value in an error case was err, a variable that was never initialized to
>>> anything other than 0.
>>>
>>> The following patch rewrites the above code to first select a string based
>>> on various conditions, and then to copy it into a newly allocated memory
>>> region, using kasprintf. This decreases subtantially the code size
>>> and removes the memory leak. The instruction for getting the length of the
>>> string and the associated variable declaration are also deleted.
>>>
>>> The patch also drops err, changes the return value to retval, which in each
>>> file was already initialized elsewhere to an error code, and initializes
>>> retval to -ENOMEM when kasprintf fails.
>>>
>>> The semantic patch that motivated this transformation is:
>>> (http://coccinelle.lip6.fr/)
>>>
>>> // <smpl>
>>> @@
>>> expression a,flag,len;
>>> expression arg,e1,e2;
>>> statement S;
>>> @@
>>>
>>> len = strlen(arg)
>>> ... when != len = e1
>>> when != arg = e2
>>> a =
>>> - \(kmalloc\|kzalloc\)(len+1,flag)
>>> + kasprintf(flag,"%s",arg)
>>> <... when != a
>>> if (a == NULL || ...) S
>>> ...>
>>> - memcpy(a,arg,len+1);
>>> // </smpl>
>>>
>>> Signed-off-by: Julia Lawall <julia@...u.dk>
>>>
>>> ---
>>> This patch makes quite a lot of changes and is only compile tested.
>>>
>>> drivers/staging/cx25821/cx25821-audio-upstream.c | 36 +++---------
>>> drivers/staging/cx25821/cx25821-video-upstream-ch2.c | 57 +++++++------------
>>> drivers/staging/cx25821/cx25821-video-upstream.c | 56 +++++++-----------
>>> 3 files changed, 56 insertions(+), 93 deletions(-)
>>>
>>> diff --git a/drivers/staging/cx25821/cx25821-audio-upstream.c b/drivers/staging/cx25821/cx25821-audio-upstream.c
>>> index 27087db..a8f6343 100644
>>> --- a/drivers/staging/cx25821/cx25821-audio-upstream.c
>>> +++ b/drivers/staging/cx25821/cx25821-audio-upstream.c
>>> @@ -723,8 +723,7 @@ int cx25821_audio_upstream_init(struct cx25821_dev *dev, int channel_select)
>>> {
>>> struct sram_channel *sram_ch;
>>> int retval = 0;
>>> - int err = 0;
>>> - int str_length = 0;
>>> + char *filename;
>>>
>>> if (dev->_audio_is_running) {
>>> printk(KERN_WARNING "Audio Channel is still running so return!\n");
>>> @@ -752,28 +751,15 @@ int cx25821_audio_upstream_init(struct cx25821_dev *dev, int channel_select)
>>> dev->_audio_lines_count = LINES_PER_AUDIO_BUFFER;
>>> _line_size = AUDIO_LINE_SIZE;
>>>
>>> - if (dev->input_audiofilename) {
>>> - str_length = strlen(dev->input_audiofilename);
>>> - dev->_audiofilename = kmalloc(str_length + 1, GFP_KERNEL);
>>> -
>>> - if (!dev->_audiofilename)
>>> - goto error;
>>> -
>>> - memcpy(dev->_audiofilename, dev->input_audiofilename,
>>> - str_length + 1);
>>> -
>>> - /* Default if filename is empty string */
>>> - if (strcmp(dev->input_audiofilename, "") == 0)
>>> - dev->_audiofilename = "/root/audioGOOD.wav";
>>> -
>>> - } else {
>>> - str_length = strlen(_defaultAudioName);
>>> - dev->_audiofilename = kmalloc(str_length + 1, GFP_KERNEL);
>>> -
>>> - if (!dev->_audiofilename)
>>> - goto error;
>>> -
>>> - memcpy(dev->_audiofilename, _defaultAudioName, str_length + 1);
>>> + if (dev->input_audiofilename &&
>>> + strcmp(dev->input_audiofilename, "") != 0)
>>> + filename = dev->input_audiofilename;
>>> + else
>>> + filename = _defaultAudioName;
>>> + dev->_audiofilename = kasprintf(GFP_KERNEL, "%s", filename);
>>> + if (!dev->_audiofilename) {
>>> + retval = -ENOMEM;
>>> + goto error;
>>> }
>>>
>> Is filename needed here at all ?
>
> I'm not sure to understand the question. There indeed appear to be
> different options. The kasprintf (which seems like it could be changed to
> kstrdup) could indeed be moved into the two branches, but it seemed nicer
> to have only one function call and one block of error handling code, since
> the error handling code is identical.
the long version is:
the variable char * filename seems to be used as tmp buffer for the result of
the if statement. IMHO you could use dev->_audiofilename directly without loss
of clarity.
Turning the if condition on its head (and removing the strcmp) is the way i would
check. Please see this a comment the result is of cause identical.
hope that helps
re
wh
>
>> The if statement looks strange this looks more familar to me:
>> if (!dev->input_audiofilename || *dev->input_audiofilename==0)
>> filename = _defaultAudioName;
>> else
>> filename = dev->input_audiofilename;
>
> OK.
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists