From owner-svn-src-all@freebsd.org Sat Aug 4 12:14:23 2018 Return-Path: Delivered-To: svn-src-all@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id E7DF31068B93; Sat, 4 Aug 2018 12:14:22 +0000 (UTC) (envelope-from jhb@FreeBSD.org) Received: from mail.baldwin.cx (bigwig.baldwin.cx [96.47.65.170]) (using TLSv1 with cipher DHE-RSA-CAMELLIA256-SHA (256/256 bits)) (Client did not present a certificate) by mx1.freebsd.org (Postfix) with ESMTPS id 8E9A372803; Sat, 4 Aug 2018 12:14:22 +0000 (UTC) (envelope-from jhb@FreeBSD.org) Received: from John-Baldwins-MacBook-Pro-2.local (unknown [IPv6:2601:648:8880:1e30:219f:c5f:b5d2:8746]) by mail.baldwin.cx (Postfix) with ESMTPSA id B4D7110AFCD; Sat, 4 Aug 2018 08:14:20 -0400 (EDT) Subject: Re: svn commit: r337273 - head/sys/dev/nvme To: Konstantin Belousov , Justin Hibbits References: <201808032004.w73K46XJ053249@repo.freebsd.org> <20180804080840.GI6049@kib.kiev.ua> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org From: John Baldwin Message-ID: Date: Sat, 4 Aug 2018 05:14:31 -0700 User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.12; rv:52.0) Gecko/20100101 Thunderbird/52.8.0 MIME-Version: 1.0 In-Reply-To: <20180804080840.GI6049@kib.kiev.ua> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit X-Greylist: Sender succeeded SMTP AUTH, not delayed by milter-greylist-4.4.3 (mail.baldwin.cx); Sat, 04 Aug 2018 08:14:21 -0400 (EDT) X-Virus-Scanned: clamav-milter 0.99.2 at mail.baldwin.cx X-Virus-Status: Clean X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.27 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 04 Aug 2018 12:14:23 -0000 On 8/4/18 1:08 AM, Konstantin Belousov wrote: > On Fri, Aug 03, 2018 at 08:04:06PM +0000, Justin Hibbits wrote: >> Author: jhibbits >> Date: Fri Aug 3 20:04:06 2018 >> New Revision: 337273 >> URL: https://svnweb.freebsd.org/changeset/base/337273 >> >> Log: >> nvme(4): Add bus_dmamap_sync() at the end of the request path >> >> Summary: >> Some architectures, in this case powerpc64, need explicit synchronization >> barriers vs device accesses. >> >> Prior to this change, when running 'make buildworld -j72' on a 18-core >> (72-thread) POWER9, I would see controller resets often. With this change, I >> don't see these resets messages, though another tester still does, for yet to be >> determined reasons, so this may not be a complete fix. Additionally, I see a >> ~5-10% speed up in buildworld times, likely due to not needing to reset the >> controller. >> >> Reviewed By: jimharris >> Differential Revision: https://reviews.freebsd.org/D16570 >> >> Modified: >> head/sys/dev/nvme/nvme_qpair.c >> >> Modified: head/sys/dev/nvme/nvme_qpair.c >> ============================================================================== >> --- head/sys/dev/nvme/nvme_qpair.c Fri Aug 3 19:24:04 2018 (r337272) >> +++ head/sys/dev/nvme/nvme_qpair.c Fri Aug 3 20:04:06 2018 (r337273) >> @@ -401,9 +401,13 @@ nvme_qpair_complete_tracker(struct nvme_qpair *qpair, >> req->retries++; >> nvme_qpair_submit_tracker(qpair, tr); >> } else { >> - if (req->type != NVME_REQUEST_NULL) >> + if (req->type != NVME_REQUEST_NULL) { >> + bus_dmamap_sync(qpair->dma_tag_payload, >> + tr->payload_dma_map, >> + BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); >> bus_dmamap_unload(qpair->dma_tag_payload, >> tr->payload_dma_map); >> + } >> >> nvme_free_request(req); >> tr->req = NULL; >> @@ -487,6 +491,8 @@ nvme_qpair_process_completions(struct nvme_qpair *qpai >> */ >> return (false); >> >> + bus_dmamap_sync(qpair->dma_tag, qpair->queuemem_map, >> + BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); >> while (1) { >> cpl = qpair->cpl[qpair->cq_head]; >> >> @@ -828,7 +834,16 @@ nvme_qpair_submit_tracker(struct nvme_qpair *qpair, st >> if (++qpair->sq_tail == qpair->num_entries) >> qpair->sq_tail = 0; >> >> + bus_dmamap_sync(qpair->dma_tag, qpair->queuemem_map, >> + BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); >> +#ifndef __powerpc__ >> + /* >> + * powerpc's bus_dmamap_sync() already includes a heavyweight sync, but >> + * no other archs do. >> + */ >> wmb(); >> +#endif > What is the purpose of this call ? It is useless without paired read > barrier. So where is the reciprocal rmb() ? For DMA, the rmb is in the device controller. However, architectures that need this kind of ordering should do it in their bus_dmmap_sync op, and this explicit one needs to be removed. (Alpha had a wmb in its bus_dmamap_sync op for this reason.) -- John Baldwin