From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paul Mackerras Subject: Re: [PATCH 16/18] KVM: PPC: Book3S HV: Add transactional memory support Date: Mon, 24 Mar 2014 12:18:22 +1100 Message-ID: <20140324011822.GB15593@iris.ozlabs.ibm.com> References: <1389176736-26821-1-git-send-email-paulus@samba.org> <1389176736-26821-17-git-send-email-paulus@samba.org> <2B37D4F2-36E4-48BD-987C-1068E644B52B@suse.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: kvm-devel , kvm-ppc To: Alexander Graf Return-path: Content-Disposition: inline In-Reply-To: <2B37D4F2-36E4-48BD-987C-1068E644B52B@suse.de> Sender: kvm-ppc-owner@vger.kernel.org List-Id: kvm.vger.kernel.org On Mon, Jan 27, 2014 at 01:47:30PM +0100, Alexander Graf wrote: > > On 08.01.2014, at 11:25, Paul Mackerras wrote: > > > + /* TM checkpointed state is now setup. All GPRs are now volatile. */ > > + TRECHKPT > > Where do we fault to if we get a transaction abort now? We can get one because we're not in suspended TM state, right? We're in non-transactional state before the trechkpt. instruction and in suspended state after it, so in neither case will a transaction abort cause a jump back to the tbegin. > > -<<<<<<< HEAD > > -======= > > Ah, so here you remove it again ;). Please run a bisectability test on the next patch set: Oops, sorry, yes I'll do that. Paul. From mboxrd@z Thu Jan 1 00:00:00 1970 From: Paul Mackerras Date: Mon, 24 Mar 2014 01:18:22 +0000 Subject: Re: [PATCH 16/18] KVM: PPC: Book3S HV: Add transactional memory support Message-Id: <20140324011822.GB15593@iris.ozlabs.ibm.com> List-Id: References: <1389176736-26821-1-git-send-email-paulus@samba.org> <1389176736-26821-17-git-send-email-paulus@samba.org> <2B37D4F2-36E4-48BD-987C-1068E644B52B@suse.de> In-Reply-To: <2B37D4F2-36E4-48BD-987C-1068E644B52B@suse.de> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Alexander Graf Cc: kvm-devel , kvm-ppc On Mon, Jan 27, 2014 at 01:47:30PM +0100, Alexander Graf wrote: > > On 08.01.2014, at 11:25, Paul Mackerras wrote: > > > + /* TM checkpointed state is now setup. All GPRs are now volatile. */ > > + TRECHKPT > > Where do we fault to if we get a transaction abort now? We can get one because we're not in suspended TM state, right? We're in non-transactional state before the trechkpt. instruction and in suspended state after it, so in neither case will a transaction abort cause a jump back to the tbegin. > > -<<<<<<< HEAD > > -===> > Ah, so here you remove it again ;). Please run a bisectability test on the next patch set: Oops, sorry, yes I'll do that. Paul.