Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH] net: stmmac: Don't call _irqoff() with hardirqs enabled
To:     John Keeping <john@metanate.com>, netdev@vger.kernel.org
Cc:     Giuseppe Cavallaro <peppe.cavallaro@st.com>,
        Alexandre Torgue <alexandre.torgue@st.com>,
        Jose Abreu <joabreu@synopsys.com>,
        "David S. Miller" <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>,
        Maxime Coquelin <mcoquelin.stm32@gmail.com>,
        linux-stm32@st-md-mailman.stormreply.com,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org
References: <20201008162749.860521-1-john@metanate.com>
From:   Heiner Kallweit <hkallweit1@gmail.com>
Message-ID: <8036d473-68bd-7ee7-e2e9-677ff4060bd3@gmail.com>
Date:   Fri, 9 Oct 2020 16:54:06 +0200
User-Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:68.0) Gecko/20100101
 Thunderbird/68.12.1
MIME-Version: 1.0
In-Reply-To: <20201008162749.860521-1-john@metanate.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 08.10.2020 18:27, John Keeping wrote:
> With threadirqs, stmmac_interrupt() is called on a thread with hardirqs
> enabled so we cannot call __napi_schedule_irqoff().  Under lockdep it
> leads to:
> 
> 	------------[ cut here ]------------
> 	WARNING: CPU: 0 PID: 285 at kernel/softirq.c:598 __raise_softirq_irqoff+0x6c/0x1c8
> 	IRQs not disabled as expected
> 	Modules linked in: brcmfmac hci_uart btbcm cfg80211 brcmutil
> 	CPU: 0 PID: 285 Comm: irq/41-eth0 Not tainted 5.4.69-rt39 #1
> 	Hardware name: Rockchip (Device Tree)
> 	[<c0110d3c>] (unwind_backtrace) from [<c010c284>] (show_stack+0x10/0x14)
> 	[<c010c284>] (show_stack) from [<c0855504>] (dump_stack+0xa8/0xe0)
> 	[<c0855504>] (dump_stack) from [<c0120a9c>] (__warn+0xe0/0xfc)
> 	[<c0120a9c>] (__warn) from [<c0120e80>] (warn_slowpath_fmt+0x7c/0xa4)
> 	[<c0120e80>] (warn_slowpath_fmt) from [<c01278c8>] (__raise_softirq_irqoff+0x6c/0x1c8)
> 	[<c01278c8>] (__raise_softirq_irqoff) from [<c056bccc>] (stmmac_interrupt+0x388/0x4e0)
> 	[<c056bccc>] (stmmac_interrupt) from [<c0178714>] (irq_forced_thread_fn+0x28/0x64)
> 	[<c0178714>] (irq_forced_thread_fn) from [<c0178924>] (irq_thread+0x124/0x260)
> 	[<c0178924>] (irq_thread) from [<c0142ee8>] (kthread+0x154/0x164)
> 	[<c0142ee8>] (kthread) from [<c01010bc>] (ret_from_fork+0x14/0x38)
> 	Exception stack(0xeb7b5fb0 to 0xeb7b5ff8)
> 	5fa0:                                     00000000 00000000 00000000 00000000
> 	5fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> 	5fe0: 00000000 00000000 00000000 00000000 00000013 00000000
> 	irq event stamp: 48
> 	hardirqs last  enabled at (50): [<c085c200>] prb_unlock+0x7c/0x8c
> 	hardirqs last disabled at (51): [<c085c0dc>] prb_lock+0x58/0x100
> 	softirqs last  enabled at (0): [<c011e770>] copy_process+0x550/0x1654
> 	softirqs last disabled at (25): [<c01786ec>] irq_forced_thread_fn+0x0/0x64
> 	---[ end trace 0000000000000002 ]---
> 
> Use __napi_schedule() instead which will save & restore the interrupt
> state.
> 
I'm thinking about a __napi_schedule version that disables hard irq's
conditionally, based on variable force_irqthreads, exported by the irq
subsystem. This would allow to behave correctly with threadirqs set,
whilst not loosing the _irqoff benefit with threadirqs unset.
Let me come up with a proposal.