netdev - Re: [PATCH V7 net-next 02/11] net/mlx5: Implement cqe_compress

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

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <4dde7c9e-92ac-43b4-b5c8-a60c92849878@gmail.com>
Date: Mon, 20 Oct 2025 08:24:19 +0300
From: Tariq Toukan <ttoukan.linux@...il.com>
To: Daniel Zahka <daniel.zahka@...il.com>, Saeed Mahameed <saeed@...nel.org>,
 "David S. Miller" <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>,
 Paolo Abeni <pabeni@...hat.com>, Eric Dumazet <edumazet@...gle.com>
Cc: Saeed Mahameed <saeedm@...dia.com>, netdev@...r.kernel.org,
 Tariq Toukan <tariqt@...dia.com>, Gal Pressman <gal@...dia.com>,
 Leon Romanovsky <leonro@...dia.com>, Jiri Pirko <jiri@...dia.com>,
 Jacob Keller <jacob.e.keller@...el.com>, Simon Horman <horms@...nel.org>
Subject: Re: [PATCH V7 net-next 02/11] net/mlx5: Implement cqe_compress_type
 via devlink params



On 18/10/2025 0:54, Daniel Zahka wrote:
> 
> 
> On 10/17/25 5:51 PM, Daniel Zahka wrote:
>>
>>
>> On 9/6/25 9:29 PM, Saeed Mahameed wrote:
>>> From: Saeed Mahameed <saeedm@...dia.com>
>>>
>>> Selects which algorithm should be used by the NIC in order to decide 
>>> rate of
>>> CQE compression dependeng on PCIe bus conditions.
>>>
>>> Supported values:
>>>
>>> 1) balanced, merges fewer CQEs, resulting in a moderate compression 
>>> ratio
>>>     but maintaining a balance between bandwidth savings and performance
>>> 2) aggressive, merges more CQEs into a single entry, achieving a higher
>>>     compression rate and maximizing performance, particularly under high
>>>     traffic loads.
>>>
>>
>> Hello,
>>
>> I'm facing some issues when trying to use the devlink param introduced 
>> in this patch. I have a multihost system with two hosts per CX7.
>>
>> My NIC is:
>> $ lshw -C net
>>   *-network
>>        description: Ethernet interface
>>        product: MT2910 Family [ConnectX-7]
>>        vendor: Mellanox Technologies
>>
>> My fw version is: 28.43.1014
>>
>> To reproduce the problem I simply read the current cqe_compress_type 
>> setting and then change it:
>>
>> $ devlink dev param show pci/0000:56:00.0 name cqe_compress_type
>> pci/0000:56:00.0:
>>   name cqe_compress_type type driver-specific
>>     values:
>>       cmode permanent value balanced
>>
>> $ devlink dev param set pci/0000:56:00.0 name cqe_compress_type value 
>> "aggressive" cmode permanent
>> kernel answers: Connection timed out
>>
>> from dmesg:
>> [  257.111349] mlx5_core 0000:56:00.0: 
>> wait_func_handle_exec_timeout:1159:(pid 72061): cmd[0]: 
>> ACCESS_REG(0x805) No done completion
>> [  257.137072] mlx5_core 0000:56:00.0: wait_func:1189:(pid 72061): 
>> ACCESS_REG(0x805) timeout. Will cause a leak of a command resource
>> [  270.871521] mlx5_core 0000:56:00.0: mlx5_cmd_comp_handler:1709:(pid 
>> 0): Command completion arrived after timeout (entry idx = 0).
>>
>>
>> subsequent attempts to use mstfwreset hang:
>>
>> $ ./mstfwreset -y -d 56:00.0 reset
>> -E- Failed to send Register MFRL: Timed out trying to take the ICMD 
>> semaphore (520).
>>
>> I can toggle the parameter ok using the mstconfig binary built from 
>> the mstflint github repo.
>>
>> Let me know if I can provide any more information.
>>
>> Daniel
> 
> Sorry, I should have mentioned my kernel version. It is a vanilla net- 
> next kernel from:
> 1c51450f1aff ("tcp: better handle TCP_TX_DELAY on established flows")
> 
> 

Hi Daniel,

Thanks for your report.
We'll look into it and reply shortly.

Regards,
Tariq