From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga02.intel.com (mga02.intel.com [134.134.136.20]) by dpdk.org (Postfix) with ESMTP id DA5BB10BD for ; Mon, 30 Apr 2018 13:21:46 +0200 (CEST) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga002.fm.intel.com ([10.253.24.26]) by orsmga101.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 30 Apr 2018 04:21:44 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.49,346,1520924400"; d="scan'208";a="41760915" Received: from irvmail001.ir.intel.com ([163.33.26.43]) by fmsmga002.fm.intel.com with ESMTP; 30 Apr 2018 04:21:43 -0700 Received: from sivswdev01.ir.intel.com (sivswdev01.ir.intel.com [10.237.217.45]) by irvmail001.ir.intel.com (8.14.3/8.13.6/MailSET/Hub) with ESMTP id w3UBLhf7016599; Mon, 30 Apr 2018 12:21:43 +0100 Received: from sivswdev01.ir.intel.com (localhost [127.0.0.1]) by sivswdev01.ir.intel.com with ESMTP id w3UBLh4K022785; Mon, 30 Apr 2018 12:21:43 +0100 Received: (from aburakov@localhost) by sivswdev01.ir.intel.com with LOCAL id w3UBLhjO022781; Mon, 30 Apr 2018 12:21:43 +0100 From: Anatoly Burakov To: dev@dpdk.org Cc: bruce.richardson@intel.com Date: Mon, 30 Apr 2018 12:21:43 +0100 Message-Id: <73e35ca9ce462ab097c6ca8e7cbfdd5b8bc06c1d.1525086045.git.anatoly.burakov@intel.com> X-Mailer: git-send-email 1.7.0.7 In-Reply-To: References: In-Reply-To: References: Subject: [dpdk-dev] [PATCH v2 2/2] mem: unmap unneeded space X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Mon, 30 Apr 2018 11:21:47 -0000 When we ask to reserve virtual areas, we usually include alignment in the mapping size, and that memory ends up being wasted. Wasting a gigabyte of VA space while trying to reserve one gigabyte is pretty expensive on 32-bit, so after we're done mapping, unmap unneeded space. Signed-off-by: Anatoly Burakov --- Notes: v2: - Split fix for size_t overflow into separate patch - Improve readability of unmapping code - Added comment explaining why unmapping is done lib/librte_eal/common/eal_common_memory.c | 26 +++++++++++++++++++++++++- 1 file changed, 25 insertions(+), 1 deletion(-) diff --git a/lib/librte_eal/common/eal_common_memory.c b/lib/librte_eal/common/eal_common_memory.c index 0ac7b33..60aed4a 100644 --- a/lib/librte_eal/common/eal_common_memory.c +++ b/lib/librte_eal/common/eal_common_memory.c @@ -121,8 +121,32 @@ eal_get_virtual_area(void *requested_addr, size_t *size, RTE_LOG(DEBUG, EAL, "Virtual area found at %p (size = 0x%zx)\n", aligned_addr, *size); - if (unmap) + if (unmap) { munmap(mapped_addr, map_sz); + } else if (!no_align) { + void *map_end, *aligned_end; + size_t before_len, after_len; + + /* when we reserve space with alignment, we add alignment to + * mapping size. On 32-bit, if 1GB alignment was requested, this + * would waste 1GB of address space, which is a luxury we cannot + * afford. so, if alignment was performed, check if any unneeded + * address space can be unmapped back. + */ + + map_end = RTE_PTR_ADD(mapped_addr, (size_t)map_sz); + aligned_end = RTE_PTR_ADD(aligned_addr, *size); + + /* unmap space before aligned mmap address */ + before_len = RTE_PTR_DIFF(aligned_addr, mapped_addr); + if (before_len > 0) + munmap(mapped_addr, before_len); + + /* unmap space after aligned end mmap address */ + after_len = RTE_PTR_DIFF(map_end, aligned_end); + if (after_len > 0) + munmap(aligned_end, after_len); + } baseaddr_offset += *size; -- 2.7.4