From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from bombadil.infradead.org ([65.50.211.133]:40292 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751919AbdFSPGD (ORCPT ); Mon, 19 Jun 2017 11:06:03 -0400 Date: Mon, 19 Jun 2017 08:05:47 -0700 From: Christoph Hellwig To: Kishon Vijay Abraham I Cc: Christoph Hellwig , linux-pci@vger.kernel.org Subject: Re: PCI endpoint API question Message-ID: <20170619150547.GA22227@infradead.org> References: <20170618140530.GA22519@infradead.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: Sender: linux-pci-owner@vger.kernel.org List-ID: On Mon, Jun 19, 2017 at 03:20:29PM +0530, Kishon Vijay Abraham I wrote: > > for writes to BARs - is this and intentional omission? It seems > > like the only current option is to poll for changes. > > You mean write to the address that's mapped to BAR in the EP side? > > PCI doesn't allow hosts to interrupt the EP (AFAIK), so we have to poll for any > writes by the host to EP memory. At least out on the bus each MMIO access translates to PCIe Memory Read/Write TLPs, so the EP itself for sure gets a notification. I don't really know enough about existing programmable endpoint IP to know how that could be forwarded to software, though.