From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bob Picco Date: Thu, 25 Sep 2014 18:54:38 +0000 Subject: Re: [PATCH V3] sparc64: sparse irq Message-Id: <20140925185438.GF4211@zareason> List-Id: References: <1410982288-10549-1-git-send-email-bpicco@meloft.net> In-Reply-To: <1410982288-10549-1-git-send-email-bpicco@meloft.net> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: sparclinux@vger.kernel.org David Miller wrote: [Thu Sep 25 2014, 01:34:20PM EDT] > From: Bob Picco > Date: Thu, 25 Sep 2014 13:14:18 -0400 > > > David Miller wrote: [Thu Sep 25 2014, 12:35:20PM EDT] > >> From: Bob Picco > >> Date: Wed, 17 Sep 2014 15:31:28 -0400 > >> > >> Major version 2 of HV_GRP_INTR was meant to behave the way that > >> you say major version 3 does. > > This is what I recall but could be incorrect. > > I an attempt to test this, I'm using your patch but I modified it > to negotiate HV_GRP_INTR major=2 and then use the cookie only > routines if that succeeded. > > The negotiation of major=2 works, but hv calls fail with EINVAL and > other strange things. > > Trying to investigate further, because making this work with major=2 > would be great because we could avoid the ivector table allocation > on even more machines. Now that I understand your motivation. Okay. I just looked quickly at the current specification for UltraSPARC Virtualization which possibly isn't publicly available. For Group HV_GRP_INTR 1.X and 3.X are supported. 2.X was declared broken. This I didn't have accessible to me at the time of debug. Also one needs to read with extreme caution.