From: Avi Kivity <avi@redhat.com>
To: qemu-devel@nongnu.org
Subject: [Qemu-devel] [PATCH 05/20] memory: change memory registration to rebuild the memory map on each change
Date: Tue, 14 Feb 2012 11:27:35 +0200 [thread overview]
Message-ID: <1329211670-11548-6-git-send-email-avi@redhat.com> (raw)
In-Reply-To: <1329211670-11548-1-git-send-email-avi@redhat.com>
Instead of incrementally building the memory map, rebuild it every time.
This allows later simplification, since the code need not consider overlaying
a previous mapping. It is also RCU friendly.
With large memory guests this can get expensive, since the operation is
O(mem size), but this will be optimized later.
As a side effect subpage and L2 leaks are fixed here.
Signed-off-by: Avi Kivity <avi@redhat.com>
---
exec.c | 50 +++++++++++++++++++++++++++++++++++++++++++++++++-
1 files changed, 49 insertions(+), 1 deletions(-)
diff --git a/exec.c b/exec.c
index 6726afd..b36c301 100644
--- a/exec.c
+++ b/exec.c
@@ -2520,6 +2520,53 @@ static subpage_t *subpage_init (target_phys_addr_t base, ram_addr_t *phys,
} \
} while (0)
+static void destroy_page_desc(PhysPageDesc pd)
+{
+ unsigned io_index = pd.phys_offset & ~TARGET_PAGE_MASK;
+ MemoryRegion *mr = io_mem_region[io_index];
+
+ if (mr->subpage) {
+ subpage_t *subpage = container_of(mr, subpage_t, iomem);
+ memory_region_destroy(&subpage->iomem);
+ g_free(subpage);
+ }
+}
+
+static void destroy_l2_mapping(void **lp, unsigned level)
+{
+ unsigned i;
+ void **p;
+ PhysPageDesc *pd;
+
+ if (!*lp) {
+ return;
+ }
+
+ if (level > 0) {
+ p = *lp;
+ for (i = 0; i < L2_SIZE; ++i) {
+ destroy_l2_mapping(&p[i], level - 1);
+ }
+ g_free(p);
+ } else {
+ pd = *lp;
+ for (i = 0; i < L2_SIZE; ++i) {
+ destroy_page_desc(pd[i]);
+ }
+ g_free(pd);
+ }
+ *lp = NULL;
+}
+
+static void destroy_all_mappings(void)
+{
+ unsigned i;
+
+ for (i = 0; i < P_L1_SIZE; ++i) {
+ destroy_l2_mapping(&l1_phys_map[i], P_L1_SHIFT / L2_BITS - 1);
+ }
+}
+
/* register physical memory.
For RAM, 'size' must be a multiple of the target page size.
If (phys_offset & ~TARGET_PAGE_MASK) != 0, then it is an
@@ -3490,6 +3537,7 @@ static void io_mem_init(void)
static void core_begin(MemoryListener *listener)
{
+ destroy_all_mappings();
}
static void core_commit(MemoryListener *listener)
@@ -3505,12 +3553,12 @@ static void core_region_add(MemoryListener *listener,
static void core_region_del(MemoryListener *listener,
MemoryRegionSection *section)
{
- cpu_register_physical_memory_log(section, false);
}
static void core_region_nop(MemoryListener *listener,
MemoryRegionSection *section)
{
+ cpu_register_physical_memory_log(section, section->readonly);
}
static void core_log_start(MemoryListener *listener,
--
1.7.9
next prev parent reply other threads:[~2012-02-14 9:28 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-02-14 9:27 [Qemu-devel] [PATCH 00/20] Reduce storage overhead of memory core Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 01/20] memory: allow MemoryListeners to observe a specific address space Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 02/20] xen: ignore I/O memory regions Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 03/20] memory: split memory listener for the two address spaces Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 04/20] memory: support stateless memory listeners Avi Kivity
2012-02-14 9:27 ` Avi Kivity [this message]
2012-02-14 9:27 ` [Qemu-devel] [PATCH 06/20] memory: remove first level of l1_phys_map Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 07/20] memory: unify phys_map last level with intermediate levels Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 08/20] memory: store MemoryRegionSection pointers in phys_map Avi Kivity
2012-03-07 17:49 ` Peter Maydell
2012-03-07 19:32 ` Peter Maydell
2012-03-08 9:50 ` Avi Kivity
2012-03-08 10:09 ` Peter Maydell
2012-03-08 11:11 ` Avi Kivity
2012-03-08 11:25 ` Peter Maydell
2012-02-14 9:27 ` [Qemu-devel] [PATCH 09/20] memory: compress phys_map node pointers to 16 bits Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 10/20] memory: fix RAM subpages in newly initialized pages Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 11/20] memory: unify the two branches of cpu_register_physical_memory_log() Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 12/20] memory: move tlb flush to MemoryListener commit callback Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 13/20] memory: make phys_page_find() return a MemoryRegionSection Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 14/20] memory: give phys_page_find() its own tree search loop Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 15/20] memory: simplify multipage/subpage registration Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 16/20] memory: replace phys_page_find_alloc() with phys_page_set() Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 17/20] memory: switch phys_page_set() to a recursive implementation Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 18/20] memory: change phys_page_set() to set multiple pages Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 19/20] memory: unify PhysPageEntry::node and ::leaf Avi Kivity
2012-02-14 9:27 ` [Qemu-devel] [PATCH 20/20] memory: allow phys_map tree paths to terminate early Avi Kivity
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=1329211670-11548-6-git-send-email-avi@redhat.com \
--to=avi@redhat.com \
--cc=qemu-devel@nongnu.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).