From mboxrd@z Thu Jan 1 00:00:00 1970 From: sboyd@codeaurora.org (Stephen Boyd) Date: Wed, 20 Aug 2014 12:11:14 -0700 Subject: [PATCH v4] irqchip: gic: Allow gic_arch_extn hooks to call into scheduler In-Reply-To: References: <1407938238-21413-1-git-send-email-sboyd@codeaurora.org> <20140817173236.GG12769@titan.lakedaemon.net> <20140817185522.GX30401@n2100.arm.linux.org.uk> <20140817190434.GJ12769@titan.lakedaemon.net> <20140817214123.GY30401@n2100.arm.linux.org.uk> Message-ID: <53F4F2D2.2050600@codeaurora.org> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 08/17/14 17:17, Nicolas Pitre wrote: > On Sun, 17 Aug 2014, Russell King - ARM Linux wrote: > >> I have no problem with changing gic_raise_softirq() to use a different >> lock, which gic_migrate_target(), and gic_set_affinity() can also use. >> There's no need for horrid locking here, because the only thing we're >> protecting is gic_map[] and the write to the register to trigger an >> IPI - and nothing using gic_arch_extn has any business knowing about >> SGIs. >> >> No need for these crappy sgi_map_lock() macros and all the ifdeffery. > Those macros are there only to conditionalize the locking in > gic_raise_softirq() because no locking what so ever is needed there when > gic_migrate_target() is configured out. I suggested the macros to cut > down on the #ifdefery in the code. Ok I can resend with the sgi lock around the gic_cpu_map updating code. Let's see how v5 goes. -- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation