Please queue up f4dca95fc0f6 for 6.9 et.al.

Please queue up f4dca95fc0f6 for 6.9 et.al.

[Holger Hoffstätte]

Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
has been queued up.

Thanks!
Holger

Re: Please queue up f4dca95fc0f6 for 6.9 et.al.

[Greg KH]

On Tue, Jun 04, 2024 at 04:56:24PM +0200, Holger Hoffstätte wrote:
> 
> Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
> commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
> has been queued up.

Only applies to 6.9.y, have backports for older kernels?

thanks,

greg k-h

Re: Please queue up f4dca95fc0f6 for 6.9 et.al.

[Holger Hoffstätte]

On 2024-06-04 17:44, Greg KH wrote:
> On Tue, Jun 04, 2024 at 04:56:24PM +0200, Holger Hoffstätte wrote:
>>
>> Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
>> commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
>> has been queued up.
> 
> Only applies to 6.9.y, have backports for older kernels?

No, sorry - I'm just the messenger here and moved everything to 6.9 already.
Cc'ing Jakub and Eric.

My understanding is that the previous commit was a performance enhancement,
so if this turns out to be too difficult then maybe 378979e94e95 ("tcp: remove
64 KByte limit for initial tp->rcv_wnd value") should just not be merged either.
I have both patches on 6.9 but really cannot say whether they should go to
older releases.

Let's wait for Jakub or Eric.

cheers
Holger

Re: Please queue up f4dca95fc0f6 for 6.9 et.al.

[Eric Dumazet]

On Tue, Jun 4, 2024 at 6:26 PM Holger Hoffstätte
<holger@applied-asynchrony.com> wrote:
>
> On 2024-06-04 17:44, Greg KH wrote:
> > On Tue, Jun 04, 2024 at 04:56:24PM +0200, Holger Hoffstätte wrote:
> >>
> >> Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
> >> commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
> >> has been queued up.
> >
> > Only applies to 6.9.y, have backports for older kernels?
>
> No, sorry - I'm just the messenger here and moved everything to 6.9 already.
> Cc'ing Jakub and Eric.
>
> My understanding is that the previous commit was a performance enhancement,
> so if this turns out to be too difficult then maybe 378979e94e95 ("tcp: remove
> 64 KByte limit for initial tp->rcv_wnd value") should just not be merged either.
> I have both patches on 6.9 but really cannot say whether they should go to
> older releases.
>

Sorry I am missing the prior emails, 378979e94e95 does not seem
security related to me,
only one small TCP change.

What is the problem ?

Re: Please queue up f4dca95fc0f6 for 6.9 et.al.

[Eric Dumazet]

On Tue, Jun 4, 2024 at 6:32 PM Eric Dumazet <edumazet@google.com> wrote:
>
> On Tue, Jun 4, 2024 at 6:26 PM Holger Hoffstätte
> <holger@applied-asynchrony.com> wrote:
> >
> > On 2024-06-04 17:44, Greg KH wrote:
> > > On Tue, Jun 04, 2024 at 04:56:24PM +0200, Holger Hoffstätte wrote:
> > >>
> > >> Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
> > >> commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
> > >> has been queued up.
> > >
> > > Only applies to 6.9.y, have backports for older kernels?
> >
> > No, sorry - I'm just the messenger here and moved everything to 6.9 already.
> > Cc'ing Jakub and Eric.
> >
> > My understanding is that the previous commit was a performance enhancement,
> > so if this turns out to be too difficult then maybe 378979e94e95 ("tcp: remove
> > 64 KByte limit for initial tp->rcv_wnd value") should just not be merged either.
> > I have both patches on 6.9 but really cannot say whether they should go to
> > older releases.
> >
>
> Sorry I am missing the prior emails, 378979e94e95 does not seem
> security related to me,
> only one small TCP change.
>
> What is the problem ?

Ah, I guess you are referring to

commit f4dca95fc0f6350918f2e6727e35b41f7f86fcce
Author: Eric Dumazet <edumazet@google.com>
Date:   Thu May 23 13:05:27 2024 +0000

    tcp: reduce accepted window in NEW_SYN_RECV state


Sure, If a stable kernel got 378979e94e95, it also needs
commit f4dca95fc0f6350918

Re: Please queue up f4dca95fc0f6 for 6.9 et.al.

[Holger Hoffstätte]

On 2024-06-04 18:32, Eric Dumazet wrote:
> On Tue, Jun 4, 2024 at 6:26 PM Holger Hoffstätte
> <holger@applied-asynchrony.com> wrote:
>>
>> On 2024-06-04 17:44, Greg KH wrote:
>>> On Tue, Jun 04, 2024 at 04:56:24PM +0200, Holger Hoffstätte wrote:
>>>>
>>>> Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
>>>> commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
>>>> has been queued up.
>>>
>>> Only applies to 6.9.y, have backports for older kernels?
>>
>> No, sorry - I'm just the messenger here and moved everything to 6.9 already.
>> Cc'ing Jakub and Eric.
>>
>> My understanding is that the previous commit was a performance enhancement,
>> so if this turns out to be too difficult then maybe 378979e94e95 ("tcp: remove
>> 64 KByte limit for initial tp->rcv_wnd value") should just not be merged either.
>> I have both patches on 6.9 but really cannot say whether they should go to
>> older releases.
>>
> 
> Sorry I am missing the prior emails, 378979e94e95 does not seem
> security related to me,
> only one small TCP change.
> 
> What is the problem ?

I noticed that 378979e94e95 was queued up for the next -stable releases
(autobot?) and notified Greg that the followup fix f4dca95fc0f6 would be a
good idea as well (since *that* one seemed quite security related to me?).
But it does not apply cleanly on older releases and I don't really feel
qualified to do backports.
So the question was whether only the first patch is OK by itself or neither
should go into older releases at all.

Hope that explains it?

Thanks!
Holger

Re: Please queue up f4dca95fc0f6 for 6.9 et.al.

[Holger Hoffstätte]

On 2024-06-04 18:35, Eric Dumazet wrote:
> On Tue, Jun 4, 2024 at 6:32 PM Eric Dumazet <edumazet@google.com> wrote:
>>
>> On Tue, Jun 4, 2024 at 6:26 PM Holger Hoffstätte
>> <holger@applied-asynchrony.com> wrote:
>>>
>>> On 2024-06-04 17:44, Greg KH wrote:
>>>> On Tue, Jun 04, 2024 at 04:56:24PM +0200, Holger Hoffstätte wrote:
>>>>>
>>>>> Just ${Subject} since it's a fix for a potential security footgun/DOS, whereever
>>>>> commit 378979e94e95 ("tcp: remove 64 KByte limit for initial tp->rcv_wnd value")
>>>>> has been queued up.
>>>>
>>>> Only applies to 6.9.y, have backports for older kernels?
>>>
>>> No, sorry - I'm just the messenger here and moved everything to 6.9 already.
>>> Cc'ing Jakub and Eric.
>>>
>>> My understanding is that the previous commit was a performance enhancement,
>>> so if this turns out to be too difficult then maybe 378979e94e95 ("tcp: remove
>>> 64 KByte limit for initial tp->rcv_wnd value") should just not be merged either.
>>> I have both patches on 6.9 but really cannot say whether they should go to
>>> older releases.
>>>
>>
>> Sorry I am missing the prior emails, 378979e94e95 does not seem
>> security related to me,
>> only one small TCP change.
>>
>> What is the problem ?
> 
> Ah, I guess you are referring to
> 
> commit f4dca95fc0f6350918f2e6727e35b41f7f86fcce
> Author: Eric Dumazet <edumazet@google.com>
> Date:   Thu May 23 13:05:27 2024 +0000
> 
>      tcp: reduce accepted window in NEW_SYN_RECV state
> 
> 
> Sure, If a stable kernel got 378979e94e95, it also needs
> commit f4dca95fc0f6350918

So I guess then it's probaby best to just drop them both from everwhere
and be done with it.

thanks,
Holger