netdev - Re: [PATCH net v3 2/4] net: txgbe: remove separate irq request for MSI and INTx

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

Message-ID: <ZoVTJIGmBMP4gCD3@localhost.localdomain>
Date: Wed, 3 Jul 2024 15:33:24 +0200
From: Michal Kubiak <michal.kubiak@...el.com>
To: Jiawen Wu <jiawenwu@...stnetic.com>
CC: <davem@...emloft.net>, <edumazet@...gle.com>, <kuba@...nel.org>,
	<pabeni@...hat.com>, <horms@...nel.org>, <andrew@...n.ch>,
	<netdev@...r.kernel.org>, <przemyslaw.kitszel@...el.com>,
	<mengyuanlou@...-swift.com>, <duanqiangwen@...-swift.com>
Subject: Re: [PATCH net v3 2/4] net: txgbe: remove separate irq request for
 MSI and INTx

On Mon, Jul 01, 2024 at 03:14:14PM +0800, Jiawen Wu wrote:
> When using MSI or INTx interrupts, request_irq() for pdev->irq will
> conflict with request_threaded_irq() for txgbe->misc.irq, to cause
> system crash. So remove txgbe_request_irq() for MSI/INTx case, and
> rename txgbe_request_msix_irqs() since it only request for queue irqs.
> 

Could you please provide any test scenario how the crash can be
reproduced and/or add an example crash log?
Also, could you describe how request_irq() conflicts with
request_threaded_irq() in this case?

> Add wx->misc_irq_domain to determine whether the driver creates an IRQ
> domain and threaded request the IRQs.
> 
> Fixes: aefd013624a1 ("net: txgbe: use irq_domain for interrupt controller")
> Signed-off-by: Jiawen Wu <jiawenwu@...stnetic.com>
> ---
>  drivers/net/ethernet/wangxun/libwx/wx_hw.c    |  1 +
>  drivers/net/ethernet/wangxun/libwx/wx_lib.c   |  5 +-
>  drivers/net/ethernet/wangxun/libwx/wx_type.h  |  1 +
>  .../net/ethernet/wangxun/txgbe/txgbe_irq.c    | 80 ++-----------------
>  .../net/ethernet/wangxun/txgbe/txgbe_irq.h    |  2 +-
>  .../net/ethernet/wangxun/txgbe/txgbe_main.c   |  2 +-
>  6 files changed, 15 insertions(+), 76 deletions(-)
> 
> diff --git a/drivers/net/ethernet/wangxun/libwx/wx_hw.c b/drivers/net/ethernet/wangxun/libwx/wx_hw.c
> index 7c4b6881a93f..d1b682ce9c6d 100644
> --- a/drivers/net/ethernet/wangxun/libwx/wx_hw.c
> +++ b/drivers/net/ethernet/wangxun/libwx/wx_hw.c
> @@ -1959,6 +1959,7 @@ int wx_sw_init(struct wx *wx)
>  	}
>  
>  	bitmap_zero(wx->state, WX_STATE_NBITS);
> +	wx->misc_irq_domain = false;
>  
>  	return 0;
>  }
> diff --git a/drivers/net/ethernet/wangxun/libwx/wx_lib.c b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
> index f53776877f71..e1f514b21090 100644
> --- a/drivers/net/ethernet/wangxun/libwx/wx_lib.c
> +++ b/drivers/net/ethernet/wangxun/libwx/wx_lib.c
> @@ -1997,7 +1997,8 @@ void wx_free_irq(struct wx *wx)
>  	int vector;
>  
>  	if (!(pdev->msix_enabled)) {
> -		free_irq(pdev->irq, wx);
> +		if (!wx->misc_irq_domain)
> +			free_irq(pdev->irq, wx);

Does it mean "pdev->irq" will never be freed if you set misc_irq_domain
to "true"? It seems you set it to true always during the initializaion,
in "txgbe_setup_misc_irq()"?

>  		return;
>  	}
>  
> @@ -2012,7 +2013,7 @@ void wx_free_irq(struct wx *wx)
>  		free_irq(entry->vector, q_vector);
>  	}
>  
> -	if (wx->mac.type == wx_mac_em)
> +	if (!wx->misc_irq_domain)
>  		free_irq(wx->msix_entry->vector, wx);
>  }
>  EXPORT_SYMBOL(wx_free_irq);
> diff --git a/drivers/net/ethernet/wangxun/libwx/wx_type.h b/drivers/net/ethernet/wangxun/libwx/wx_type.h
> index 5aaf7b1fa2db..0df7f5712b6f 100644
> --- a/drivers/net/ethernet/wangxun/libwx/wx_type.h
> +++ b/drivers/net/ethernet/wangxun/libwx/wx_type.h
> @@ -1058,6 +1058,7 @@ struct wx {
>  	dma_addr_t isb_dma;
>  	u32 *isb_mem;
>  	u32 isb_tag[WX_ISB_MAX];
> +	bool misc_irq_domain;

I don't think that introducing a kind of global variable to determine the call
context is a good idea.
Also, it seems that member is always true after the "probe" context is
completed, isn't it?

> 

[...]

> @@ -256,6 +190,8 @@ int txgbe_setup_misc_irq(struct txgbe *txgbe)
>  	if (err)
>  		goto free_gpio_irq;
>  
> +	wx->misc_irq_domain = true;
> +
>  	return 0;
> 

Is there any chance that member will be set back to "false" after the
initialization is completed?


Thanks,
Michal