mbox series

[f/j/m:linux-aws,0/1] tcp: Add memory barrier to tcp_push

Message ID 20240222182903.1490015-1-philip.cox@canonical.com
Headers show
Series tcp: Add memory barrier to tcp_push | expand

Message

Philip Cox Feb. 22, 2024, 6:29 p.m. UTC 
BugLink: https://bugs.launchpad.net/bugs/2051727

SRU Justification:

[Impact]

Recently a patch has been accepted on the Linux Kernel
regarding the TCP stack which fix an additional +40ms
latency on ARM64 CPU architecture without impacting
other CPU types.

[Fix]

Upstream patch:
   7267e8dcad6b2f9fce05a6a06335d7040acbc2b6: tcp: Add memory barrier to tcp_push()
was submitted to address this.

[Test Plan]

AWS tested

[Where problems could occur]

The risk of regressions from this change is fairly low.

--


Salvatore Dipietro (1):
  tcp: Add memory barrier to tcp_push()

 net/ipv4/tcp.c | 1 +
 1 file changed, 1 insertion(+)

Comments

Tim Gardner Feb. 23, 2024, 3:24 p.m. UTC | #1 
On 2/22/24 11:29, Philip Cox wrote:
> 
> BugLink: https://bugs.launchpad.net/bugs/2051727
> 
> SRU Justification:
> 
> [Impact]
> 
> Recently a patch has been accepted on the Linux Kernel
> regarding the TCP stack which fix an additional +40ms
> latency on ARM64 CPU architecture without impacting
> other CPU types.
> 
> [Fix]
> 
> Upstream patch:
>     7267e8dcad6b2f9fce05a6a06335d7040acbc2b6: tcp: Add memory barrier to tcp_push()
> was submitted to address this.
> 
> [Test Plan]
> 
> AWS tested
> 
> [Where problems could occur]
> 
> The risk of regressions from this change is fairly low.
> 
> --
> 
> 
> Salvatore Dipietro (1):
>    tcp: Add memory barrier to tcp_push()
> 
>   net/ipv4/tcp.c | 1 +
>   1 file changed, 1 insertion(+)
> 
Acked-by: Tim Gardner <tim.gardner@canonical.com>

I happen to know that this LP bug is associated with a Sales Force case. 
Its no secret that we use SF to deal with paying customers, so I always 
add the SF case number to '[Other Info]' so that the relationship 
between LP and SF is easily traceable.

The other more compulsive observation that applies to many patch 
submissions is the discussion under '[Where problems could occur]'. That 
section doesn't need to describe the likelihood  of regression. That is 
really the purview of the reviewers. It quite literally expects a 
discussion of adverse symptoms a user might encounter. In this case it 
should be a comment about the possibility of network stalls or other 
adverse throughput impacts.
Magali Lemes March 1, 2024, 1:39 p.m. UTC | #2 
On 22/02/2024 15:29, Philip Cox wrote:
> 
> BugLink: https://bugs.launchpad.net/bugs/2051727
> 
> SRU Justification:
> 
> [Impact]
> 
> Recently a patch has been accepted on the Linux Kernel
> regarding the TCP stack which fix an additional +40ms
> latency on ARM64 CPU architecture without impacting
> other CPU types.
> 
> [Fix]
> 
> Upstream patch:
>     7267e8dcad6b2f9fce05a6a06335d7040acbc2b6: tcp: Add memory barrier to tcp_push()
> was submitted to address this.
> 
> [Test Plan]
> 
> AWS tested
> 
> [Where problems could occur]
> 
> The risk of regressions from this change is fairly low.
Acked-by: Magali Lemes <magali.lemes@canonical.com>

> 
> --
> 
> 
> Salvatore Dipietro (1):
>    tcp: Add memory barrier to tcp_push()
> 
>   net/ipv4/tcp.c | 1 +
>   1 file changed, 1 insertion(+)
>
Philip Cox March 1, 2024, 6:20 p.m. UTC | #3 
On Thu, Feb 22, 2024 at 1:31 PM Philip Cox <philip.cox@canonical.com> wrote:

>
> BugLink: https://bugs.launchpad.net/bugs/2051727
>
> SRU Justification:
>
> [Impact]
>
> Recently a patch has been accepted on the Linux Kernel
> regarding the TCP stack which fix an additional +40ms
> latency on ARM64 CPU architecture without impacting
> other CPU types.
>
> [Fix]
>
> Upstream patch:
>    7267e8dcad6b2f9fce05a6a06335d7040acbc2b6: tcp: Add memory barrier to
> tcp_push()
> was submitted to address this.
>
> [Test Plan]
>
> AWS tested
>
> [Where problems could occur]
>
> The risk of regressions from this change is fairly low.
>
> --
>
>
> Salvatore Dipietro (1):
>   tcp: Add memory barrier to tcp_push()
>
>  net/ipv4/tcp.c | 1 +
>  1 file changed, 1 insertion(+)
>
> --
> 2.34.1
>
>