Commit f980d055a0f858d73d9467bb0b570721bbfcdfb8 causes a regression

Commit f980d055a0f858d73d9467bb0b570721bbfcdfb8 causes a regression

[Tim Gardner]

Hi Len,

I have a report (https://bugs.launchpad.net/bugs/1952094) that commit 
f980d055a0f858d73d9467bb0b570721bbfcdfb8 ("CIFS: Fix a potencially 
linear read overflow") causes a regression as a stable backport in a 5.4 
based kernel. I don't know if this regression exists in tip as well, or 
if it is unique to the backported environment. I suspect, given the 
content of the patch, that it is generic. As such, it has been 
backported to a number of stable releases:

linux-4.4.y.txt:0955df2d9bf4857e3e2287e3028903e6cec06c30
linux-4.9.y.txt:8878af780747f498551b7d360cae61b415798f18
linux-4.14.y.txt:20967547ffc6039f17c63a1c24eb779ee166b245
linux-4.19.y.txt:bea655491daf39f1934a71bf576bf3499092d3a4
linux-5.4.y.txt:b444064a0e0ef64491b8739a9ae05a952b5f8974
linux-5.10.y.txt:6c4857203ffa36918136756a889b12c5864bc4ad
linux-5.13.y.txt:9bffe470e9b537075345406512df01ca2188b725
linux-5.14.y.txt:c41dd61c86482ab34f6f039b13296308018fd99b

Could this be an off-by-one issue if the source string is full length ?

rtg
-- 
-----------
Tim Gardner
Canonical, Inc

Re: Commit f980d055a0f858d73d9467bb0b570721bbfcdfb8 causes a regression

[Jeff Layton]

On Mon, 2021-11-29 at 08:35 -0700, Tim Gardner wrote:
> Hi Len,
> 
> I have a report (https://bugs.launchpad.net/bugs/1952094) that commit 
> f980d055a0f858d73d9467bb0b570721bbfcdfb8 ("CIFS: Fix a potencially 
> linear read overflow") causes a regression as a stable backport in a 5.4 
> based kernel. I don't know if this regression exists in tip as well, or 
> if it is unique to the backported environment. I suspect, given the 
> content of the patch, that it is generic. As such, it has been 
> backported to a number of stable releases:
> 
> linux-4.4.y.txt:0955df2d9bf4857e3e2287e3028903e6cec06c30
> linux-4.9.y.txt:8878af780747f498551b7d360cae61b415798f18
> linux-4.14.y.txt:20967547ffc6039f17c63a1c24eb779ee166b245
> linux-4.19.y.txt:bea655491daf39f1934a71bf576bf3499092d3a4
> linux-5.4.y.txt:b444064a0e0ef64491b8739a9ae05a952b5f8974
> linux-5.10.y.txt:6c4857203ffa36918136756a889b12c5864bc4ad
> linux-5.13.y.txt:9bffe470e9b537075345406512df01ca2188b725
> linux-5.14.y.txt:c41dd61c86482ab34f6f039b13296308018fd99b
> 
> Could this be an off-by-one issue if the source string is full length ?
> 
> rtg

Maybe? But it doesn't seem to be that long. The error message evidently
says:

    "CIFS VFS: CIFS mount error: iocharset utf8 not found"

The iocharset string ("utf8" here) usually gets set in the mount string
and then we just pass that string to load_nls().

The patch you're pointing out though doesn't seem to be involved in any
of that. It sounds like something else is wrong. I'd validate that that
patch was applied correctly, and get more details about what this guy is
doing.

g/l!
-- 
Jeff Layton <jlayton@kernel.org>

Re: Commit f980d055a0f858d73d9467bb0b570721bbfcdfb8 causes a regression

[Tim Gardner]

On 11/29/21 9:01 AM, Jeff Layton wrote:
> On Mon, 2021-11-29 at 08:35 -0700, Tim Gardner wrote:
>> Hi Len,
>>
>> I have a report (https://bugs.launchpad.net/bugs/1952094) that commit
>> f980d055a0f858d73d9467bb0b570721bbfcdfb8 ("CIFS: Fix a potencially
>> linear read overflow") causes a regression as a stable backport in a 5.4
>> based kernel. I don't know if this regression exists in tip as well, or
>> if it is unique to the backported environment. I suspect, given the
>> content of the patch, that it is generic. As such, it has been
>> backported to a number of stable releases:
>>
>> linux-4.4.y.txt:0955df2d9bf4857e3e2287e3028903e6cec06c30
>> linux-4.9.y.txt:8878af780747f498551b7d360cae61b415798f18
>> linux-4.14.y.txt:20967547ffc6039f17c63a1c24eb779ee166b245
>> linux-4.19.y.txt:bea655491daf39f1934a71bf576bf3499092d3a4
>> linux-5.4.y.txt:b444064a0e0ef64491b8739a9ae05a952b5f8974
>> linux-5.10.y.txt:6c4857203ffa36918136756a889b12c5864bc4ad
>> linux-5.13.y.txt:9bffe470e9b537075345406512df01ca2188b725
>> linux-5.14.y.txt:c41dd61c86482ab34f6f039b13296308018fd99b
>>
>> Could this be an off-by-one issue if the source string is full length ?
>>
>> rtg
> 
> Maybe? But it doesn't seem to be that long. The error message evidently
> says:
> 
>      "CIFS VFS: CIFS mount error: iocharset utf8 not found"
> 
> The iocharset string ("utf8" here) usually gets set in the mount string
> and then we just pass that string to load_nls().
> 
> The patch you're pointing out though doesn't seem to be involved in any
> of that. It sounds like something else is wrong. I'd validate that that
> patch was applied correctly, and get more details about what this guy is
> doing.
> 
> g/l!
> 

You may be correct. I made the rash assumption the reporter had done a 
normal full upgrade. However, it appears he may have only done targeted 
package updates and neglected the linux-modules-extra package that 
contains the nls_utf8 module. My test kernel installation instructions 
_did_ include linux-modules-extra which is likely why it worked for him.

I've requested more info from the reporter to verify that the missing 
package is the root of his problem.

Thanks for responding.

rtg
-----------
Tim Gardner
Canonical, Inc