netdev - Re: [PATCH v2 net] r8169: fix NAPI handling under high load

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]

Date:   Thu, 18 Oct 2018 20:17:15 +0200
From:   Heiner Kallweit <hkallweit1@...il.com>
To:     Eric Dumazet <eric.dumazet@...il.com>,
        David Miller <davem@...emloft.net>,
        Realtek linux nic maintainers <nic_swsd@...ltek.com>
Cc:     "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH v2 net] r8169: fix NAPI handling under high load

On 18.10.2018 20:02, Eric Dumazet wrote:
> 
> 
> On 10/18/2018 10:56 AM, Heiner Kallweit wrote:
>> rtl_rx() and rtl_tx() are called only if the respective bits are set
>> in the interrupt status register. Under high load NAPI may not be
>> able to process all data (work_done == budget) and it will schedule
>> subsequent calls to the poll callback.
>> rtl_ack_events() however resets the bits in the interrupt status
>> register, therefore subsequent calls to rtl8169_poll() won't call
>> rtl_rx() and rtl_tx() - chip interrupts are still disabled.
>>
>> Fix this by calling rtl_rx() and rtl_tx() independent of the bits
>> set in the interrupt status register. Both functions will detect
>> if there's nothing to do for them.
>>
>> Fixes: da78dbff2e05 ("r8169: remove work from irq handler.")
>> Signed-off-by: Heiner Kallweit <hkallweit1@...il.com>
>> ---
>> v2:
>> - added "Fixes" tag
>> ---
>>  drivers/net/ethernet/realtek/r8169.c | 8 +++-----
>>  1 file changed, 3 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
>> index 8c4f49adc..7caf3b7e9 100644
>> --- a/drivers/net/ethernet/realtek/r8169.c
>> +++ b/drivers/net/ethernet/realtek/r8169.c
>> @@ -6528,17 +6528,15 @@ static int rtl8169_poll(struct napi_struct *napi, int budget)
>>  	struct rtl8169_private *tp = container_of(napi, struct rtl8169_private, napi);
>>  	struct net_device *dev = tp->dev;
>>  	u16 enable_mask = RTL_EVENT_NAPI | tp->event_slow;
>> -	int work_done= 0;
>> +	int work_done;
>>  	u16 status;
>>  
>>  	status = rtl_get_events(tp);
>>  	rtl_ack_events(tp, status & ~tp->event_slow);
>>  
>> -	if (status & RTL_EVENT_NAPI_RX)
>> -		work_done = rtl_rx(dev, tp, (u32) budget);
>> +	work_done = rtl_rx(dev, tp, (u32) budget);
>>  
>> -	if (status & RTL_EVENT_NAPI_TX)
>> -		rtl_tx(dev, tp);
>> +	rtl_tx(dev, tp);
>>  
>>  	if (status & tp->event_slow) {
>>  		enable_mask &= ~tp->event_slow;
>>
> 
> One has to wonder why rtl8169_poll(), which might be called in a loop under DOS,
> has to call rtl_ack_events() ?
> 
> Should not this be called one time from hard irq handler instead ?
> 
Yes, it should. I have a patch in my tree including this, in addition it
moves everything from rtl_slow_event_work() to the hard irq handler.
However, due to recent changes, this patch may not apply cleanly to
older kernel versions.
Calling rtl_ack_events() in the poll callback isn't nice but not directly
a bug, so I'd prefer to submit this to net-next.

>