From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx1.redhat.com (mx1.redhat.com [209.132.183.28]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by ml01.01.org (Postfix) with ESMTPS id F057321A13493 for ; Wed, 3 May 2017 16:36:39 -0700 (PDT) Received: from smtp.corp.redhat.com (int-mx01.intmail.prod.int.phx2.redhat.com [10.5.11.11]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 5673883F45; Wed, 3 May 2017 23:36:39 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mx1.redhat.com 5673883F45 Authentication-Results: ext-mx03.extmail.prod.ext.phx2.redhat.com; dmarc=none (p=none dis=none) header.from=redhat.com Authentication-Results: ext-mx03.extmail.prod.ext.phx2.redhat.com; spf=pass smtp.mailfrom=lersek@redhat.com DKIM-Filter: OpenDKIM Filter v2.11.0 mx1.redhat.com 5673883F45 Received: from lacos-laptop-7.usersys.redhat.com (ovpn-116-118.phx2.redhat.com [10.3.116.118]) by smtp.corp.redhat.com (Postfix) with ESMTP id 7C13760466; Wed, 3 May 2017 23:36:35 +0000 (UTC) To: Gerd Hoffmann References: <1382eb04-9646-133b-9ce5-8293cb54745f@redhat.com> <1493794647.8581.144.camel@redhat.com> <159c4eae-4e13-7958-59f4-dfab4c1bf16e@redhat.com> <1493819062.8581.177.camel@redhat.com> <64591d6f-b5d9-d73d-26a5-4c157b9bd541@redhat.com> Cc: "Kinney, Michael D" , "Fan, Jeff" , "Yao, Jiewen" , edk2-devel-01 , Paolo Bonzini From: Laszlo Ersek Message-ID: Date: Thu, 4 May 2017 01:36:34 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0 MIME-Version: 1.0 In-Reply-To: <64591d6f-b5d9-d73d-26a5-4c157b9bd541@redhat.com> X-Scanned-By: MIMEDefang 2.79 on 10.5.11.11 X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.27]); Wed, 03 May 2017 23:36:39 +0000 (UTC) Subject: Re: SMRAM sizes on large hosts X-BeenThere: edk2-devel@lists.01.org X-Mailman-Version: 2.1.22 Precedence: list List-Id: EDK II Development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 03 May 2017 23:36:40 -0000 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit On 05/04/17 00:33, Laszlo Ersek wrote: > On 05/03/17 15:44, Gerd Hoffmann wrote: >> Hi, >> >>> I propose the following: add a new fw_cfg file which communicates how >>> much memory (how many megabytes) the "11b" value in the tseg size >>> register will configure. >> >> I'd prefer to keep fw_cfg out of the picture, and I think we can do it >> without too much problems. >> >> We have a TSEGMB (tseg memory base) register. qemu doesn't implement >> it (but I think you can write to it and read back what you've >> written). > > OVMF already uses this register, for the exact purpose you mention. > > (1) OVMF sets the register in SmmAccessPeiEntryPoint(), which is the > entry point of the PEI-phase module that produces PEI_SMM_ACCESS_PPI. > > (2) OVMF reads the register back in SmramAccessGetCapabilities(), which > is built into two modules: > > - the same PEIM as mentioned above, for producing > PEI_SMM_ACCESS_PPI.GetCapabilities(), > > - a similar-purpose DXE driver, for producing > EFI_SMM_ACCESS2_PROTOCOL.GetCapabilities(). > > Those GetCapabilities() member functions basically return the SMRAM map > to the caller. > >> We could make qemu update that register in case "11b" is written to >> the tseg size. The firmware then can read TSEGMB and figure whenever >> a large tseg is supported and if so how big it is. > > Do you mean that QEMU would at once calculate the absolute base of TSEG, > downwards from the top of low RAM, expressed in MB? > > Excerpt from (1): > >> // >> // Set TSEG Memory Base. >> // >> PciWrite32 (DRAMC_REGISTER_Q35 (MCH_TSEGMB), >> (TopOfLowRamMb - FixedPcdGet8 (PcdQ35TsegMbytes)) << MCH_TSEGMB_MB_SHIFT); >> >> // >> // Set TSEG size, and disable TSEG visibility outside of SMM. Note that the >> // T_EN bit has inverse meaning; when T_EN is set, then TSEG visibility is >> // *restricted* to SMM. >> // >> EsmramcVal &= ~(UINT32)MCH_ESMRAMC_TSEG_MASK; >> EsmramcVal |= FixedPcdGet8 (PcdQ35TsegMbytes) == 8 ? MCH_ESMRAMC_TSEG_8MB : >> FixedPcdGet8 (PcdQ35TsegMbytes) == 2 ? MCH_ESMRAMC_TSEG_2MB : >> MCH_ESMRAMC_TSEG_1MB; >> EsmramcVal |= MCH_ESMRAMC_T_EN; >> PciWrite8 (DRAMC_REGISTER_Q35 (MCH_ESMRAMC), EsmramcVal); > > I guess I could update this to: > - set MCH_TSEGMB like now, but also remember the written value > - write 11b to MCH_ESMRAMC, and read back MCH_TSEGMB > - if the read back value differs from the written one, use that value > for tseg size going forward, plus write 11b | T_EN to MCH_ESMRAMC > - otherwise, set MCH_ESMRAMC like now. > > The problem with this is that I need the TSEG size in another module as > well, namely PlatformPei. The dispatch order between PlatformPei and > SmmAccessPei is unspecified (they have both TRUE for DEPEX). If > PlatformPei gets dispatched second, I really wouldn't want to write to > MCH_ESMRAMC again, just to query MCH_TSEGMB. (I couldn't just read > MCH_TSEGMB because if PlatformPei were dispatched first, then MCH_TSEGMB > would still be unset). > > In other words, I wouldn't like to write a register just to receive the > information; I need the information in two places whose relative > ordering is unspecified, and one of those already writes the register in > question, genuinely. I guess I could hack it around with another dynamic > PCD, but that's kind of ugly. > > If we invented a read-only, side-effect-free PCI config space register > that gave me this value plain and simple (similarly to how a new fw_cfg > file would do), that would be a lot cleaner for me. I think this would > match your earlier alternative where you wrote "Alternatively we could > add some qemu-specific register". That's my next preference after > fw_cfg. If we invent such a new register, it should be in a location that is either read-only, or zeroed-on-reset, in current QEMU. Otherwise, new firmware running on old QEMU could be misled by a guest OS that writes to this register, and then either reboots or enters S3. ... With this in mind, I don't oppose "having to write somewhere to read back the result", but then let's please make that write access as well to the same new qemu-specific register, and not to MCH_ESMRAMC. Thanks! Laszlo