[PATCH] net: stmmac: ethtool: Fixed calltrace caused by unbalanced disable_irq_wake calls

[Simon Horman]

On Tue, Jan 16, 2024 at 01:57:34PM +0800, Qiang Ma wrote:
> On Mon, 15 Jan 2024 13:42:38 +0000
> Simon Horman <horms at kernel.org> wrote:
> 
> > + Florian Fainelli <f.fainelli at gmail.com>
> > 
> > On Fri, Jan 12, 2024 at 10:12:49AM +0800, Qiang Ma wrote:
> > > We found the following dmesg calltrace when testing the GMAC NIC
> > > notebook:
> > > 
> > > [9.448656] ------------[ cut here ]------------
> > > [9.448658] Unbalanced IRQ 43 wake disable
> > > [9.448673] WARNING: CPU: 3 PID: 1083 at kernel/irq/manage.c:688
> > > irq_set_irq_wake+0xe0/0x128 [9.448717] CPU: 3 PID: 1083 Comm:
> > > ethtool Tainted: G           O      4.19 #1 [9.448773]         ...
> > > [9.448774] Call Trace:
> > > [9.448781] [<9000000000209b5c>] show_stack+0x34/0x140
> > > [9.448788] [<9000000000d52700>] dump_stack+0x98/0xd0
> > > [9.448794] [<9000000000228610>] __warn+0xa8/0x120
> > > [9.448797] [<9000000000d2fb60>] report_bug+0x98/0x130
> > > [9.448800] [<900000000020a418>] do_bp+0x248/0x2f0
> > > [9.448805] [<90000000002035f4>] handle_bp_int+0x4c/0x78
> > > [9.448808] [<900000000029ea40>] irq_set_irq_wake+0xe0/0x128
> > > [9.448813] [<9000000000a96a7c>] stmmac_set_wol+0x134/0x150
> > > [9.448819] [<9000000000be6ed0>] dev_ethtool+0x1368/0x2440
> > > [9.448824] [<9000000000c08350>] dev_ioctl+0x1f8/0x3e0
> > > [9.448827] [<9000000000bb2a34>] sock_ioctl+0x2a4/0x450
> > > [9.448832] [<900000000046f044>] do_vfs_ioctl+0xa4/0x738
> > > [9.448834] [<900000000046f778>] ksys_ioctl+0xa0/0xe8
> > > [9.448837] [<900000000046f7d8>] sys_ioctl+0x18/0x28
> > > [9.448840] [<9000000000211ab4>] syscall_common+0x20/0x34
> > > [9.448842] ---[ end trace 40c18d9aec863c3e ]---
> > > 
> > > Multiple disable_irq_wake() calls will keep decreasing the IRQ
> > > wake_depth, When wake_depth is 0, calling disable_irq_wake() again,
> > > will report the above calltrace.
> > > 
> > > Due to the need to appear in pairs, we cannot call
> > > disable_irq_wake() without calling enable_irq_wake(). Fix this by
> > > making sure there are no unbalanced disable_irq_wake() calls.  
> > 
> > 
> > Hi Qiang Ma,
> > 
> > This seems to be a fix, so I think it should be targeted at net:
> > 
> > 	Subject: [PATCH net] ...
> > 
> > And have a fixes tag, perhaps:
> > 
> > 	Fixes: 3172d3afa998 ("stmmac: support wake up irq from
> > external sources (v3)")
> > 
> > I don't think there is a need to repost this patch because of the
> > above, but please keep it in mind for next time.
> > 
> > > Signed-off-by: Qiang Ma <maqianga at uniontech.com>  
> > 
> > I see that the approach taken here is the same as that taken
> > by bcm_sysport_set_wol() to what seems to be a similar problem [1].
> > So the code change itself looks good to me.
> > 
> > Reviewed-by: Simon Horman <horms at kernel.org>
> > 
> > [1] 61b423a8a0bd ("net: systemport: avoid unbalanced enable_irq_wake
> > calls") https://git.kernel.org/torvalds/c/61b423a8a0bd
> > 
> > > ---
> > >  drivers/net/ethernet/stmicro/stmmac/stmmac.h         |  1 +
> > >  drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c | 10
> > > ++++++++-- drivers/net/ethernet/stmicro/stmmac/stmmac_main.c    |
> > > 1 + 3 files changed, 10 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
> > > b/drivers/net/ethernet/stmicro/stmmac/stmmac.h index
> > > cd7a9768de5f..b8c93b881a65 100644 ---
> > > a/drivers/net/ethernet/stmicro/stmmac/stmmac.h +++
> > > b/drivers/net/ethernet/stmicro/stmmac/stmmac.h @@ -255,6 +255,7 @@
> > > struct stmmac_priv { u32 msg_enable;
> > >  	int wolopts;
> > >  	int wol_irq;
> > > +	bool wol_irq_disabled;
> > >  	int clk_csr;
> > >  	struct timer_list eee_ctrl_timer;
> > >  	int lpi_irq;
> > > diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c
> > > b/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c index
> > > f628411ae4ae..9a4d9492a781 100644 ---
> > > a/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c +++
> > > b/drivers/net/ethernet/stmicro/stmmac/stmmac_ethtool.c @@ -825,10
> > > +825,16 @@ static int stmmac_set_wol(struct net_device *dev, struct
> > > ethtool_wolinfo *wol) if (wol->wolopts) { pr_info("stmmac: wakeup
> > > enable\n"); device_set_wakeup_enable(priv->device, 1);
> > > -		enable_irq_wake(priv->wol_irq);
> > > +		/* Avoid unbalanced enable_irq_wake calls */
> > > +		if (priv->wol_irq_disabled)
> > > +			enable_irq_wake(priv->wol_irq);
> > > +		priv->wol_irq_disabled = false;
> > >  	} else {
> > >  		device_set_wakeup_enable(priv->device, 0);
> > > -		disable_irq_wake(priv->wol_irq);
> > > +		/* Avoid unbalanced disable_irq_wake calls */
> > > +		if (!priv->wol_irq_disabled)
> > > +			disable_irq_wake(priv->wol_irq);
> > > +		priv->wol_irq_disabled = true;
> > >  	}
> > >  
> > >  	mutex_lock(&priv->lock);
> > > diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
> > > b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c index
> > > 37e64283f910..baa396621ed8 100644 ---
> > > a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c +++
> > > b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c @@ -3565,6
> > > +3565,7 @@ static int stmmac_request_irq_multi_msi(struct
> > > net_device *dev) /* Request the Wake IRQ in case of another line
> > >  	 * is used for WoL
> > >  	 */
> > > +	priv->wol_irq_disabled = true;
> > >  	if (priv->wol_irq > 0 && priv->wol_irq != dev->irq) {
> > >  		int_name = priv->int_name_wol;
> > >  		sprintf(int_name, "%s:%s", dev->name, "wol");
> > > -- 
> > > 2.20.1
> > >   
> > 
> 
> Hi Simon Horman,
> 
> The latest code does not seem to see the stmmac driver to avoid
> irq_wake call related fix committed, this fix is mainly for stmmac,
> refer to the commit:
> 
> commit 61b423a8a0bd9aeaa046f9a24bed42e3a953a936 Author:
> Florian Fainelli <f.fainelli at gmail.com> Date:   Fri Oct 10 10:51:54
> 2014 -0700
> 
>     net: systemport: avoid unbalanced enable_irq_wake calls
> 
> commit 083731a8fbe71d83fc908adf137dc98ee352f280
> Author: Florian Fainelli <f.fainelli at gmail.com>
> Date:   Fri Oct 10 10:51:53 2014 -0700
> 
>     net: bcmgenet: avoid unbalanced enable_irq_wake calls
> 
> Therefore, I think this submission is necessary at this time.

Yes, I agree.