From patchwork Sat Jun 24 01:24:57 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Alison Schofield X-Patchwork-Id: 13291550 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 0054FEB64DD for ; Sat, 24 Jun 2023 01:25:06 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S231367AbjFXBZG (ORCPT ); Fri, 23 Jun 2023 21:25:06 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:56432 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S231270AbjFXBZE (ORCPT ); Fri, 23 Jun 2023 21:25:04 -0400 Received: from mga17.intel.com (mga17.intel.com [192.55.52.151]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 9CBBD2736 for ; Fri, 23 Jun 2023 18:25:03 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1687569903; x=1719105903; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=IAfREYdqo1McQjjaEXB6IDj9iFefK7mPRuguP6YQVYM=; b=CPktTv/KwfJ3YyHTAliba1cpdgt20aBEuDGWHhEvh5v+qJp24G859dw9 9pOHz38mT/8fGkwpJHPSzT+3PpyL7Zr615FzOUvKq3SbDdf/6+4ihQBxA 649YrYZGLr/KkEoH4Hcb2A/AqYkqt7cU2N6uKgfVWrjcHYkH/qFo9AfYa Eu2f/dGGNWZXASL8ODkVXbRPyK5mFKY3sOxSbaSbDKTU468fg0CAS2bWA jAhi47Y6SRvbXA6OhdMjeu3bhzEk2Jekk97TmaxQUoTWoEbCFLPnsriHU cis6FIRVJv8/fjJ4/SMaBfV+PeIITK54Y76+H+IWjMQMx2/RT6bPfp4tg w==; X-IronPort-AV: E=McAfee;i="6600,9927,10750"; a="341241793" X-IronPort-AV: E=Sophos;i="6.01,153,1684825200"; d="scan'208";a="341241793" Received: from fmsmga008.fm.intel.com ([10.253.24.58]) by fmsmga107.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 23 Jun 2023 18:25:03 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10750"; a="780767800" X-IronPort-AV: E=Sophos;i="6.01,153,1684825200"; d="scan'208";a="780767800" Received: from aschofie-mobl2.amr.corp.intel.com (HELO localhost) ([10.251.28.92]) by fmsmga008-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 23 Jun 2023 18:25:02 -0700 From: alison.schofield@intel.com To: Dan Williams , Ira Weiny , Vishal Verma , Dave Jiang , Ben Widawsky Cc: Alison Schofield , linux-cxl@vger.kernel.org, Derick Marks Subject: [PATCH 2/2] cxl/region: Remove a soft reserved resource at region teardown Date: Fri, 23 Jun 2023 18:24:57 -0700 Message-Id: <1167b10afd2bfd61dc872b1f95efabf954bbb79d.1687568084.git.alison.schofield@intel.com> X-Mailer: git-send-email 2.39.2 In-Reply-To: References: MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: linux-cxl@vger.kernel.org From: Alison Schofield When CXL regions are created through autodiscovery their resource may be a child of a soft reserved resource. Today, when such a region is destroyed, its soft reserved resource remains in place. That means that the HPA range that the region could release, is actually unavailable for reuse. Free the soft reserved resource on region teardown by examining the alignment of the resources, and handling accordingly. The two resources, will be exactly aligned, partially aligned, or not aligned at all. |----------- "Soft Reserved" -----------| |-------------- "Region #" -------------| Exactly aligned. Any dangling children move up to a parent on removal. The removal of this soft reserved seems guaranteed, however, the availability of the address range for use depends on complete cleanup of the region's child resources also. |----------- "Soft Reserved" -----------| |-------- "Region #" -------| or |----------- "Soft Reserved" -----------| |-------- "Region #" -------| Either start or end aligns. Unlike removing a resource, which simply moves child resources up the resource tree, adjustments fail if any child resources map the range being truncated. So, this one will fail for dangling child resources of the region. |---------- "Soft Reserved" ----------| |---- "Region #" ----| No alignment. Freeing the resource of a region in the middle is possible if no child resources map the leading or trailing address space. Reported-by: Derick Marks Signed-off-by: Alison Schofield --- drivers/cxl/core/region.c | 132 +++++++++++++++++++++++++++++++++++--- 1 file changed, 123 insertions(+), 9 deletions(-) diff --git a/drivers/cxl/core/region.c b/drivers/cxl/core/region.c index 1769ea92bf3a..2cdb9a2770f3 100644 --- a/drivers/cxl/core/region.c +++ b/drivers/cxl/core/region.c @@ -525,22 +525,136 @@ static int alloc_hpa(struct cxl_region *cxlr, resource_size_t size) return 0; } +static int add_soft_reserved(struct resource *parent, resource_size_t start, + resource_size_t len, unsigned long flags) +{ + struct resource *res; + int rc; + + res = kmalloc(sizeof(*res), GFP_KERNEL); + if (!res) + return -ENOMEM; + + *res = DEFINE_RES_MEM_NAMED(start, len, "Soft Reserved"); + + res->desc = IORES_DESC_SOFT_RESERVED; + res->flags = res->flags | flags; + rc = insert_resource(parent, res); + if (rc) + kfree(res); + + return rc; +} + +static void remove_soft_reserved(struct cxl_region *cxlr, struct resource *soft, + resource_size_t start, resource_size_t end) +{ + struct cxl_root_decoder *cxlrd = to_cxl_root_decoder(cxlr->dev.parent); + resource_size_t new_start, new_end; + struct resource *parent; + unsigned long flags; + int rc; + + /* Prevent new usage while removing or adjusting the resource */ + mutex_lock(&cxlrd->range_lock); + + /* Aligns at both resource start and end */ + if (soft->start == start && soft->end == end) { + rc = remove_resource(soft); + if (rc) + dev_dbg(&cxlr->dev, + "cannot remove soft reserved resource %pr\n", + soft); + else + kfree(soft); + + goto out; + } + + /* Aligns at either resource start or end */ + if (soft->start == start || soft->end == end) { + if (soft->start == start) { + new_start = end + 1; + new_end = soft->end; + } else { + new_start = soft->start; + new_end = start + 1; + } + rc = adjust_resource(soft, new_start, new_end - new_start + 1); + if (rc) + dev_dbg(&cxlr->dev, + "cannot adjust soft reserved resource %pr\n", + soft); + goto out; + } + + /* + * No alignment. Attempt a 3-way split that removes the part of + * the resource the region occupied, and then creates new soft + * reserved resources for the leading and trailing addr space. + * adjust_resource() will stop the attempt if there are any + * child resources. + */ + + /* Save the original soft reserved resource params before adjusting */ + new_start = soft->start; + new_end = soft->end; + parent = soft->parent; + flags = soft->flags; + + rc = adjust_resource(soft, start, end - start); + if (rc) { + dev_dbg(&cxlr->dev, + "cannot adjust soft reserved resource %pr\n", soft); + goto out; + } + rc = remove_resource(soft); + if (rc) + dev_warn(&cxlr->dev, + "cannot remove soft reserved resource %pr\n", soft); + + rc = add_soft_reserved(parent, new_start, start - new_start, flags); + if (rc) + dev_warn(&cxlr->dev, + "cannot add new soft reserved resource at %pa\n", + &new_start); + + rc = add_soft_reserved(parent, end + 1, new_end - end, flags); + if (rc) + dev_warn(&cxlr->dev, + "cannot add new soft reserved resource at %pa + 1\n", + &end); +out: + mutex_unlock(&cxlrd->range_lock); +} + static void cxl_region_iomem_release(struct cxl_region *cxlr) { struct cxl_region_params *p = &cxlr->params; + struct resource *parent, *res = p->res; + resource_size_t start, end; if (device_is_registered(&cxlr->dev)) lockdep_assert_held_write(&cxl_region_rwsem); - if (p->res) { - /* - * Autodiscovered regions may not have been able to insert their - * resource. - */ - if (p->res->parent) - remove_resource(p->res); - kfree(p->res); - p->res = NULL; + if (!res) + return; + /* + * Autodiscovered regions may not have been able to insert their + * resource. If a Soft Reserved parent resource exists, try to + * remove that also. + */ + if (p->res->parent) { + parent = p->res->parent; + start = p->res->start; + end = p->res->end; + remove_resource(p->res); + if (test_bit(CXL_REGION_F_AUTO, &cxlr->flags) && + parent->desc == IORES_DESC_SOFT_RESERVED) + remove_soft_reserved(cxlr, parent, start, end); } + + kfree(p->res); + p->res = NULL; } static int free_hpa(struct cxl_region *cxlr)