From mboxrd@z Thu Jan 1 00:00:00 1970 Message-ID: <4FDBABEB.90300@xenomai.org> Date: Fri, 15 Jun 2012 23:40:59 +0200 From: Philippe Gerum MIME-Version: 1.0 References: In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Xenomai] x86 MSI support List-Id: Discussions about the Xenomai project List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Jeff Weber Cc: Xenomai Mailing List On 06/15/2012 08:44 PM, Jeff Weber wrote: > Is the FAQ line: > > "The I-pipe patch currently does not support MSI interrupts." > > still true? This FAQ is dated to 2008. > In retrospect, this FAQ entry is somewhat misleading. As Jan pointed out a couple of times already, the I-pipe does handle MSI interrupts properly, but doesn't allow non-root domains to call the IRQ masking/unmasking code for MSI. Since unmasking is usually paired with requesting an IRQ, and masking is paired with releasing it, the restriction should include the request/free calls as well. There is no reason to mask MSI in the pre-handling code due to their semantics, and therefore no reason to unmask them upon return from the handler, so the restriction turns out to be a non-issue for most use cases. If a primary domain handler really wants to prevent more MSI to come from a given source, then it should quiesce the issuing device. > I've been using MSIs and adeos-ipipe-2.6.38.8-x86-2.10-01.patch and haven't > seen a problem yet. > -- Philippe.