From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paolo Bonzini Subject: Re: [PATCH] kvm: x86: merge kvm_arch_set_irq and kvm_arch_set_irq_inatomic Date: Fri, 18 Nov 2016 20:20:03 +0100 Message-ID: <0e695811-5053-adfa-1ac9-3e9c55dbc596@redhat.com> References: <1479394547-15249-3-git-send-email-pbonzini@redhat.com> <4d1b0863-8c51-079a-b74e-8950c51060ff@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Transfer-Encoding: 7bit To: David Hildenbrand , linux-kernel@vger.kernel.org, kvm@vger.kernel.org Return-path: In-Reply-To: <4d1b0863-8c51-079a-b74e-8950c51060ff@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On 18/11/2016 19:20, David Hildenbrand wrote: > Am 17.11.2016 um 15:55 schrieb Paolo Bonzini: >> kvm_arch_set_irq is unused since commit b97e6de9c96. Merge >> its functionality with kvm_arch_set_irq_inatomic. >> >> Reported-by: Jiang Biao >> Signed-off-by: Paolo Bonzini >> --- > > As you said, it is unused. Therefore the functionality is superfluous. > Why merge it? Because we can handle Hyper-V synthetic interrupts atomically, and that was the intended usage of kvm_arch_set_irq's code (see commit c9a5eccac1ab, "kvm/eventfd: add arch-specific set_irq", 2015-10-16). What happened was that the API changed between commit c9a5eccac1ab and the merge of the Hyper-V synthetic interrupt patches, and the latter was not adjusted. Paolo > We can still introduce this later if we ever need it. Or do you have a > concrete user in mind? > > The patch in general looks good to me. Just wondering if we can't simply > rip out that single function. >