From patchwork Wed Feb 18 02:12:57 2009 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Subject: =?gb2312?B?tPC4tDogW1BBVENIXSBmaXggdGhlIGludGVycnVwdCBsb3NzIHBybw==?= =?gb2312?B?YmxlbSBvbiBwb3dlcnBjIElQSUMgKDIuNi4yNS0yLjYuMjgp?= Date: Tue, 17 Feb 2009 16:12:57 -0000 From: dayu@datangmobile.cn X-Patchwork-Id: 23320 Message-Id: To: , Cc: linuxppc-dev@ozlabs.org, linux-kernel@vger.kernel.org From: Da Yu Date: Wed, 18 Feb 2009 19:58:20 +0800 Subject: [PATCH] fix the interrupt loss problem on powerpc IPIC (2.6.25-2.6.28) Signed-off-by: Da Yu --- -----邮件原件----- 发件人: Li Yang [mailto:leoli@freescale.com] 发送时间: 2009年2月17日 22:38 收件人: Kumar Gala 抄送: 笪禹; linux-kernel@vger.kernel.org; linuxppc-dev@ozlabs.org 主题: Re: [PATCH] fix the interrupt loss problem on powerpc IPIC (2.6.23) On Tue, Feb 17, 2009 at 10:12 PM, Kumar Gala wrote: > > On Feb 17, 2009, at 6:44 AM, > > wrote: > >> From: Da Yu >> Date: Tue, 17 Feb 2009 19:58:20 +0800 >> Subject: [PATCH] fix the interrupt loss problem on powerpc IPIC >> (2.6.23) >> >> Signed-off-by: Da Yu >> --- > > Please provide a bit more description as to why this fixes the issue. The pending register is write 1 clear. If there are more than one external interrupts pending at the same time, acking the first interrupt will also clear other interrupt pending bits. That will cause loss of interrupt. - Leo --- a/arch/powerpc/sysdev/ipic.c 2009-02-18 09:47:04.000000000 +0800 +++ b/arch/powerpc/sysdev/ipic.c 2009-02-18 09:46:34.000000000 +0800 @@ -568,8 +568,7 @@ static void ipic_ack_irq(unsigned int vi spin_lock_irqsave(&ipic_lock, flags); - temp = ipic_read(ipic->regs, ipic_info[src].ack); - temp |= (1 << (31 - ipic_info[src].bit)); + temp = 1 << (31 - ipic_info[src].bit); ipic_write(ipic->regs, ipic_info[src].ack, temp); /* mb() can't guarantee that ack is finished. But it does finish @@ -592,8 +591,7 @@ static void ipic_mask_irq_and_ack(unsign temp &= ~(1 << (31 - ipic_info[src].bit)); ipic_write(ipic->regs, ipic_info[src].mask, temp); - temp = ipic_read(ipic->regs, ipic_info[src].ack); - temp |= (1 << (31 - ipic_info[src].bit)); + temp = 1 << (31 - ipic_info[src].bit); ipic_write(ipic->regs, ipic_info[src].ack, temp); /* mb() can't guarantee that ack is finished. But it does finish