From: Anatoly Burakov <anatoly.burakov@intel.com>
To: dev@dpdk.org
Cc: Bruce Richardson <bruce.richardson@intel.com>
Subject: [dpdk-dev] [PATCH v2 2/3] eal/bsdapp: concatenate adjacent segments
Date: Thu, 28 Jun 2018 12:41:48 +0100 [thread overview]
Message-ID: <385f43246053aaf19d2377dbb9f91dbf8c314e35.1530186042.git.anatoly.burakov@intel.com> (raw)
In-Reply-To: <f046c737e0304a8fdbf23b766577757634119f31.1530186042.git.anatoly.burakov@intel.com>
In-Reply-To: <fd5598072a48e941be2f851b6286078b50e0c7b2.1528716160.git.anatoly.burakov@intel.com>
Previously, memory allocator always left holes between mapped
contigmem segments, even if they were IOVA-contiguous. Fix this
by remembering last IOVA address and memseg index, and checking
against those when mapping new contigmem segments.
Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
---
Notes:
v2:
- Fix comparison when deciding if hole is needed
lib/librte_eal/bsdapp/eal/eal_memory.c | 48 ++++++++++++++++----------
1 file changed, 30 insertions(+), 18 deletions(-)
diff --git a/lib/librte_eal/bsdapp/eal/eal_memory.c b/lib/librte_eal/bsdapp/eal/eal_memory.c
index ca06de2f8..7bf05c760 100644
--- a/lib/librte_eal/bsdapp/eal/eal_memory.c
+++ b/lib/librte_eal/bsdapp/eal/eal_memory.c
@@ -104,6 +104,8 @@ rte_eal_hugepage_init(void)
/* map all hugepages and sort them */
for (i = 0; i < internal_config.num_hugepage_sizes; i ++){
struct hugepage_info *hpi;
+ rte_iova_t prev_end = 0;
+ int prev_ms_idx = -1;
uint64_t page_sz, mem_needed;
unsigned int n_pages, max_pages;
@@ -124,10 +126,27 @@ rte_eal_hugepage_init(void)
int error;
size_t sysctl_size = sizeof(physaddr);
char physaddr_str[64];
+ bool is_adjacent;
+
+ /* first, check if this segment is IOVA-adjacent to
+ * the previous one.
+ */
+ snprintf(physaddr_str, sizeof(physaddr_str),
+ "hw.contigmem.physaddr.%d", j);
+ error = sysctlbyname(physaddr_str, &physaddr,
+ &sysctl_size, NULL, 0);
+ if (error < 0) {
+ RTE_LOG(ERR, EAL, "Failed to get physical addr for buffer %u "
+ "from %s\n", j, hpi->hugedir);
+ return -1;
+ }
+
+ is_adjacent = prev_end != 0 && physaddr == prev_end;
+ prev_end = physaddr + hpi->hugepage_sz;
for (msl_idx = 0; msl_idx < RTE_MAX_MEMSEG_LISTS;
msl_idx++) {
- bool empty;
+ bool empty, need_hole;
msl = &mcfg->memsegs[msl_idx];
arr = &msl->memseg_arr;
@@ -136,20 +155,23 @@ rte_eal_hugepage_init(void)
empty = arr->count == 0;
- /* we need 1, plus hole if not empty */
+ /* we need a hole if this isn't an empty memseg
+ * list, and if previous segment was not
+ * adjacent to current one.
+ */
+ need_hole = !empty && !is_adjacent;
+
+ /* we need 1, plus hole if not adjacent */
ms_idx = rte_fbarray_find_next_n_free(arr,
- 0, 1 + (empty ? 1 : 0));
+ 0, 1 + (need_hole ? 1 : 0));
/* memseg list is full? */
if (ms_idx < 0)
continue;
- /* leave some space between memsegs, they are
- * not IOVA contiguous, so they shouldn't be VA
- * contiguous either.
- */
- if (!empty)
+ if (need_hole && prev_ms_idx == ms_idx - 1)
ms_idx++;
+ prev_ms_idx = ms_idx;
break;
}
@@ -178,16 +200,6 @@ rte_eal_hugepage_init(void)
return -1;
}
- snprintf(physaddr_str, sizeof(physaddr_str), "hw.contigmem"
- ".physaddr.%d", j);
- error = sysctlbyname(physaddr_str, &physaddr, &sysctl_size,
- NULL, 0);
- if (error < 0) {
- RTE_LOG(ERR, EAL, "Failed to get physical addr for buffer %u "
- "from %s\n", j, hpi->hugedir);
- return -1;
- }
-
seg->addr = addr;
seg->iova = physaddr;
seg->hugepage_sz = page_sz;
--
2.17.1
next prev parent reply other threads:[~2018-06-28 11:41 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-06-11 16:13 [dpdk-dev] [PATCH 1/3] eal/bsdapp: fix segment index display Anatoly Burakov
2018-06-11 16:13 ` [dpdk-dev] [PATCH 2/3] eal/bsdapp: concatenate adjacent segments Anatoly Burakov
2018-06-11 16:13 ` [dpdk-dev] [PATCH 3/3] eal: make memory segment preallocation OS-specific Anatoly Burakov
2018-06-28 11:41 ` [dpdk-dev] [PATCH v2 1/3] eal/bsdapp: fix segment index display Anatoly Burakov
2018-06-28 11:41 ` Anatoly Burakov [this message]
2018-06-28 11:41 ` [dpdk-dev] [PATCH v2 3/3] eal: make memory segment preallocation OS-specific Anatoly Burakov
2018-07-12 23:00 ` Thomas Monjalon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=385f43246053aaf19d2377dbb9f91dbf8c314e35.1530186042.git.anatoly.burakov@intel.com \
--to=anatoly.burakov@intel.com \
--cc=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).