From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-wr1-f50.google.com (mail-wr1-f50.google.com [209.85.221.50]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id A65D620DD51 for ; Wed, 4 Feb 2026 06:20:16 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.221.50 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1770186017; cv=none; b=f0Gn6snuNS5JbqjermPA0cUqRNAo3QTdInnHC7XOyEUHe8soNvCLRddPHsFSnvWkkMkKQGIN/lKBdYSjDLVD4X/T4fA4IafVUG/4vNkbGfY/Sv3kghzhRvdRw0SluKefxDGMk7jz9qkzfrW01gnoFSIXNYuD7zXiGBwrg/mz2x0= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1770186017; c=relaxed/simple; bh=rOWpNiQG4UmcSP34ZAePYE+b4Jm0ZqxupPtjhUQQn0Y=; h=Message-ID:Date:MIME-Version:Subject:To:Cc:References:From: In-Reply-To:Content-Type; b=Zc3i1PUAO5kQbm6HOgTHKHGop0AlE33YlKsR6m6K/apNUc/64owmN+mgvKPgxXhFH237ZGdBdCoPTPxFm/2n7Up0T/ox1TIHX45e9dlvPGtNupxOTGRKVJC/73qfeqok0oD/bLN8Hjs73NgWS7z7CkdrC7Gie1Zt2PsO8DXn9DU= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=S+T6z1Fx; arc=none smtp.client-ip=209.85.221.50 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="S+T6z1Fx" Received: by mail-wr1-f50.google.com with SMTP id ffacd0b85a97d-4359108fd24so184195f8f.2 for ; Tue, 03 Feb 2026 22:20:16 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1770186015; x=1770790815; darn=vger.kernel.org; h=content-transfer-encoding:in-reply-to:from:content-language :references:cc:to:subject:user-agent:mime-version:date:message-id :from:to:cc:subject:date:message-id:reply-to; bh=bf4hdFEzmH13IzCRYQqo8LgcKjYgw5KgSvvD+W/m134=; b=S+T6z1FxjmHAIPqR2HJGMiCJRG0S4IZ26P03bBcz489yj9X0+KxjlBizX50YIiD11S sSiLM635vmcVuVuaQIuk+e4f2Imse9YnWpc4aE3CNe5viDWZkqImGkBXBPMFtfdGXzdp ifmYIREuHhGcAQLi7mGNiU//k7yBoIJ96XO6tidKUso5TLNRYDO90fqWE9KCCEdqkfPi YBUfRSa36xo3Q0rdbOQ6J1gHr3Ceza1qvC3N+jJBYS/bdp3FMDxMQkbwnS3LJlRXpvZK +RG8TDYiUdp1rPXqCW749xOcLc8P81sVsxQhVoIE31VVDQ+yud9YaczkMrtZta+EIe4R FVJg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1770186015; x=1770790815; h=content-transfer-encoding:in-reply-to:from:content-language :references:cc:to:subject:user-agent:mime-version:date:message-id :x-gm-gg:x-gm-message-state:from:to:cc:subject:date:message-id :reply-to; bh=bf4hdFEzmH13IzCRYQqo8LgcKjYgw5KgSvvD+W/m134=; b=dEbtWmTAIIuB8e1oAio+QSFijS4jldpAKbbhg2Fzt4sxtkOLrRTSOuNY/j8sDRrUid u+A35LcZ2CFVAN2C8/Qa8kR7NHR4g92LrHbueh3DHaVe71gSuzrcArOhRnie87NioDnG CE8yQuuQDbA5hfH2q0LXqoaOSCSIqhNR96ssJOYlxJ16cK/dXgnaNe9FM5HMvLAVoMaF cGRgkaniguzNRb8eD+3bCk1+FP67pfdyfgmIViEpCJuI9yPuf6nD3JyeCIhTM6+ZCN6t vKG6O5UZeTFhFKJmXGRChGP4rwPbiCpTUzGr46ziBmvByQN80WzS8WMWFjwg70QMH1iD WqDA== X-Forwarded-Encrypted: i=1; AJvYcCV/Oxq5ZAln83tbRWeVEmFDxb+G1VrQyuHA+7eJQXbzQ75GkkUX3ef0RoOsUUCYnV+Ikw4FeKY=@vger.kernel.org X-Gm-Message-State: AOJu0Yw3Svn4vBH6ng6dTd9eC1vsRmRXqdsmPVYT49N8fAkeeImfJ8Oq oDukDnOqtT6ah1rzUmnCmWI7Z3rKjb3KGde1dZXNMveQ63k+mlMnMjQ2JYXDtA== X-Gm-Gg: AZuq6aJ8vpP0oyY/lWIeA0RtZw+tx0RDlkw/+0hzpGcaF4kPHPlFZAfNL2VOKgy6q1r OKY+ZCCgZ7Q5Zy9MKMyLXXO71oTERVKlNyEOURX0WcFB8XBtFK1tcQBGIDBwPULL6qkpRtrfKfy xLJMnjzsPq3z9+uIuxWe/RlRytLDRL7Y6KIXPYc1POgaE6pipTW+f7Vts+1l3/M2r54zIAaupAk aR+1Rd4c3SoSWgmSxuHFYhyZ2VvX7fugm2BA2hrq2xfeZc2URjSKVwoYHjasbslrvvdQJpRofJd vMkxtOJ3Dsqz282y8faIDn5r59XgVXjgAbDj3PexrNVJBaosHZjuLux4MOgbgG5d7vXvYDw+azd 18PkR884ARGBGVxZvzDVszo0V1HNPsmhZYJ8gPixY0kL4RCoOHNXnJ1w1a2/jWxe4EFrYpgQamF 5LzBrSFijd5R+5btmpXuI55bV1WPXk X-Received: by 2002:a05:6000:40d9:b0:435:a370:2d7a with SMTP id ffacd0b85a97d-43618053444mr2378362f8f.36.1770186014618; Tue, 03 Feb 2026 22:20:14 -0800 (PST) Received: from [10.80.1.200] ([72.25.96.16]) by smtp.gmail.com with ESMTPSA id ffacd0b85a97d-43618064844sm3770597f8f.42.2026.02.03.22.20.13 (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Tue, 03 Feb 2026 22:20:14 -0800 (PST) Message-ID: <8a3a3ff4-16c7-4d99-8854-38d741cc6b82@gmail.com> Date: Wed, 4 Feb 2026 08:20:14 +0200 Precedence: bulk X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 User-Agent: Mozilla Thunderbird Subject: Re: [PATCH net-next] net/mlx5e: Undo saving per-channel async ICOSQ To: Alice Mikityanska , Daniel Borkmann , netdev@vger.kernel.org Cc: William Tu , Tariq Toukan , David Wei , Jakub Kicinski , Gal Pressman References: <20260123223916.361295-1-daniel@iogearbox.net> <85776531-d5fa-4762-90aa-74c8397dc09b@gmail.com> <6188b9f5-ce38-4de9-80b7-c7b1cab48595@iogearbox.net> <421d9b09-a157-4f95-8775-7883ea097038@gmail.com> <2a3bd098-0d88-4c11-8ac3-7b13f8bbbd9f@iogearbox.net> Content-Language: en-US From: Tariq Toukan In-Reply-To: Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8bit On 03/02/2026 19:54, Alice Mikityanska wrote: > On Tue, Feb 3, 2026, at 16:58, Tariq Toukan wrote: >> On 03/02/2026 11:20, Alice Mikityanska wrote: >>> On Mon, Feb 2, 2026, at 18:13, Daniel Borkmann wrote: >>>> Hi Tariq, >>>> >>>> On 2/1/26 1:50 PM, Tariq Toukan wrote: >>>>> On 26/01/2026 11:23, Daniel Borkmann wrote: >>>>>> On 1/25/26 9:33 AM, Tariq Toukan wrote: >>>>>>> On 24/01/2026 0:39, Daniel Borkmann wrote: >>>>>>>> This reverts the following commits: >>>>>>>> >>>>>>>>    - ea945f4f3991 ("net/mlx5e: Move async ICOSQ lock into ICOSQ struct") >>>>>>>>    - 56aca3e0f730 ("net/mlx5e: Use regular ICOSQ for triggering NAPI") >>>>>>>>    - 1b080bd74840 ("net/mlx5e: Move async ICOSQ to dynamic allocation") >>>>>>>>    - abed42f9cd80 ("net/mlx5e: Conditionally create async ICOSQ") >>>>>>>> >>>>>>>> There are a couple of regressions on the xsk side I ran into: >>>>>>>> >>>>>>>> Commit 56aca3e0f730 triggers an illegal synchronize_rcu() in an RCU read- >>>>>>>> side critical section via mlx5e_xsk_wakeup() -> mlx5e_trigger_napi_icosq() >>>>>>>> -> synchronize_net(). The stack holds RCU read-lock in xsk_poll(). >>>>>>>> >>>>>>>> Additionally, this also hits a NULL pointer dereference in mlx5e_xsk_wakeup(): >>>>>>>> >>>>>>>>    [  103.963735] BUG: kernel NULL pointer dereference, address: 0000000000000240 >>>>>>>>    [  103.963743] #PF: supervisor read access in kernel mode >>>>>>>>    [  103.963746] #PF: error_code(0x0000) - not-present page >>>>>>>>    [  103.963749] PGD 0 P4D 0 >>>>>>>>    [  103.963752] Oops: Oops: 0000 [#1] SMP >>>>>>>>    [  103.963756] CPU: 0 UID: 0 PID: 2255 Comm: qemu-system-x86 Not tainted 6.19.0-rc5+ #229 PREEMPT(none) >>>>>>>>    [  103.963761] Hardware name: [...] >>>>>>>>    [  103.963765] RIP: 0010:mlx5e_xsk_wakeup+0x53/0x90 [mlx5_core] >>>>>>>> >>>>>>>> What happens is that c->async_icosq is NULL when in mlx5e_xsk_wakeup() >>>>>>>> and therefore access to c->async_icosq->state triggers it. (On the NIC >>>>>>>> there is an XDP program installed by the control plane where traffic >>>>>>>> gets redirected into an xsk map - there was no xsk pool set up yet. >>>>>>>> At some later time a xsk pool is set up and the related xsk socket is >>>>>>>> added to the xsk map of the XDP program.) >>>>>>> >>>>>>> Thanks for your report. >>>>>>> >>>>>>>> Reverting the series fixes the problems again. >>>>>>> >>>>>>> Revert is too aggressive here. A fix is preferable. >>>>>>> We're investigating the issue in order to fix it. >>>>>>> We'll update. >>>>>> Ok, sounds good. Certainly the kTLS fixes seem independent, from the cause >>>>>> of the issues I've hit it just seemed to me that they were quite fundamental >>>>>> and that perhaps a different approach would be needed (or alternatively only >>>>>> kTLS would need fixing, and the xsk optimization left as it was originally). >>>>>> Anyway, I'll keep the revert locally for now, and happy to test patches. >>>>> >>>>> Please check attached patch. >>> >>> Hi Tariq, >>> >>> I also took a glance at the patch: that seems to be correct in XSK parts, and mlx5e_trigger_napi_icosq now makes sense to me after your explanation. >> >> Great. >> >>> Two small comments though: >>> >>> 1. I'd prefer DEBUG_NET_WARN_ON_ONCE instead of WARN_ON. At least XSK wakeup is a hot path, called frequently, so we don't want to flood dmesg with the same error, and it would be better to compile out the check in non-debug builds, hence DEBUG_NET_WARN_ON_ONCE. >>> >> >> I tend to totally drop this check. >> >>> 2. I see you changed the condition of creating async_icosq to be created when any xdp_prog is attached, even when there are no XSK pools. Is there a case where async_icosq is useful with plain XDP without XSK? If not, I believe this condition should be XDP && XSK. >>> >> >> You're right. >> As you know, the channels are not re-opened on XSK pool events. >> They are re-opened on XDP program attach/detach. > > Ah, that's right, I see. I didn't pay attention to where the check is > located. If you reorganize the code a little bit, it's possible to > implement the perfect check (XDP && XSK) || kTLS_RX, to avoid creating > the unneeded async_icosq with plain XDP. > Right, but I won't do it here as a fix. > (BTW, I see that ktls_rx_was_enabled is never reset to false: is it > because we can't reliably fence the moment when the driver stops using > it?) > Offload feature can be disabled but offloaded connections can stay alive until they're closed. It is possible to track them and reset ktls_rx_was_enabled to false, but it's currently not implemented. > Anyway, it's up to you if you'd like to do it or not, but here is my > suggestion. We don't have to create async_icosq in mlx5e_open_queues. > Similar to how we separated mlx5e_open_xsk, we can separate > mlx5e_open_async_icosq, which is where we can check (XDP && XSK) || > kTLS_RX. > > Then mlx5e_open_channel will look like: > > mlx5e_open_queues(); > mlx5e_open_async_icosq(); // Checks the condition internally > if (xsk_pool) > mlx5e_open_xsk(); > > mlx5e_xsk_enable_locked will also call mlx5e_open_async_icosq just > before mlx5e_open_xsk. > > mlx5e_close_async_icosq will close it if it exists. > > With this design, all cases are covered: every mlx5e_open_xsk is paired > with mlx5e_open_async_icosq; and kTLS RX enable flow reopens channels, > which also triggers mlx5e_open_async_icosq. The conditional is DRYed in > mlx5e_open_async_icosq itself, so no code repetition either. > Right, it would work, and save the async icosq creation for the case (XDP && !XSK). I prefer to implement this change as a followup, and keep this fix simple. >> Hence, as a simple and quick fix, we will check for XDP program and >> create the async ICOSQ accordingly. >> Currently we won't have create/destroy async ICOSQ when XSK pool is >> activated/deactivated. I will mention this in the commit message. >> >> This still prevents the async ICOSQ creation in default. >> >>> Thanks, >>> Alice >>> >> >> >>>>> We were able to repro the issues internally and verify the fix. >>>>> We're finalizing it before submission. >>>>> >>>>> I'd be glad if you can confirm it solves the issues for you. >>>> That seems to work for me, yes. Feel free to add my Tested-by. >>>> >>>> Thanks, >>>> Daniel