All of lore.kernel.org
 help / color / mirror / Atom feed
From: Christian Pinto <c.pinto@virtualopensystems.com>
To: zhang.zhanghailiang@huawei.com
Cc: b.reynal@virtualopensystems.com, aarcange@redhat.com,
	quintela@redhat.com, dgilbert@redhat.com, amit.shah@redhat.com,
	peter.huangpeng@huawei.com, hanweidong@huawei.com,
	qemu-devel@nongnu.org, tech@virtualopensystems.com,
	Christian Pinto <c.pinto@virtualopensystems.com>
Subject: [Qemu-devel] [RFC PATCH 2/4] migration/ram: Fix for ARM/ARM64 page size
Date: Thu,  9 Mar 2017 12:34:35 +0100	[thread overview]
Message-ID: <20170309113437.9667-3-c.pinto@virtualopensystems.com> (raw)
In-Reply-To: <20170309113437.9667-1-c.pinto@virtualopensystems.com>

Architecture such as ARM use a page size of 1KB, while write protection
is done at the granularity of host pages (generally 4KB).
All addresses must always be aligned to the size of a host page.

Signed-off-by: Christian Pinto <c.pinto@virtualopensystems.com>
Signed-off-by: Baptiste Reynal <b.reynal@virtualopensystems.com>
---
 migration/postcopy-ram.c |  6 +++---
 migration/ram.c          | 18 ++++++++++++++----
 2 files changed, 17 insertions(+), 7 deletions(-)

diff --git a/migration/postcopy-ram.c b/migration/postcopy-ram.c
index 9c45f1059f..97382067b3 100644
--- a/migration/postcopy-ram.c
+++ b/migration/postcopy-ram.c
@@ -373,7 +373,7 @@ int postcopy_ram_prepare_discard(MigrationIncomingState *mis)
     return 0;
 }
 
-int ram_set_pages_wp(uint64_t page_addr,
+int ram_set_pages_wp(ram_addr_t page_addr,
                      uint64_t size,
                      bool remove,
                      int uffd)
@@ -556,10 +556,10 @@ static void *postcopy_ram_fault_thread(void *opaque)
             * will be an deadlock error.
             */
             if (migration_in_setup(ms)) {
-                uint64_t host = msg.arg.pagefault.address;
+                ram_addr_t host = msg.arg.pagefault.address;
 
                 host &= ~(hostpagesize - 1);
-                ret = ram_set_pages_wp(host, getpagesize(), true,
+                ret = ram_set_pages_wp(host, hostpagesize, true,
                                        us->userfault_fd);
                 if (ret < 0) {
                     error_report("Remove page's write-protect failed");
diff --git a/migration/ram.c b/migration/ram.c
index 3417c56f29..7a3b1c7ed3 100644
--- a/migration/ram.c
+++ b/migration/ram.c
@@ -1383,14 +1383,24 @@ static int ram_find_and_save_block(QEMUFile *f, bool last_stage,
             /* For snapshot, we will remove the page write-protect here */
             if (migration_in_snapshot(ms)) {
                 int ret;
-                uint64_t host_addr = (uint64_t)(pss.block->host + pss.offset);
-
+                /*
+                 * Some architectures in QEMU use a smaller memory page size
+                 * with respect to the host page size. ARM as an example
+                 * uses 1K memory pages, while Linux supports pages of minimum
+                 * size of 4K.
+                 * Userfault write protection works at the level of a host page
+                 * and thus one full host page has to be protected/unprotected
+                 * every time.
+                 */
+                ram_addr_t host_addr = (ram_addr_t)(pss.block->host +
+                                    pss.offset) & (~(qemu_host_page_size - 1));
                 ret = ram_set_pages_wp(host_addr, getpagesize(), true,
                                        ms->userfault_state.userfault_fd);
                 if (ret < 0) {
                     error_report("Failed to remove the write-protect for page:"
-                                 "%"PRIx64 " length: %d, block: %s", host_addr,
-                                 getpagesize(), pss.block->idstr);
+                                "%"PRIx64 " length: %d, offset: %"PRIx64
+                                ", block: %s", host_addr, getpagesize(),
+                                pss.offset, pss.block->idstr);
                 }
             }
         }
-- 
2.11.0

  parent reply	other threads:[~2017-03-09 11:35 UTC|newest]

Thread overview: 48+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-01-07 12:19 [Qemu-devel] [RFC 00/13] Live memory snapshot based on userfaultfd zhanghailiang
2016-01-07 12:19 ` [Qemu-devel] [RFC 01/13] postcopy/migration: Split fault related state into struct UserfaultState zhanghailiang
2016-01-07 12:19 ` [Qemu-devel] [RFC 02/13] migration: Allow the migrate command to work on file: urls zhanghailiang
2016-07-13 16:12   ` Dr. David Alan Gilbert
2016-07-14  5:27     ` Hailiang Zhang
2016-01-07 12:19 ` [Qemu-devel] [RFC 03/13] migration: Allow -incoming " zhanghailiang
2016-01-11 20:02   ` Dr. David Alan Gilbert
2016-01-12 13:04     ` Hailiang Zhang
2016-01-07 12:19 ` [Qemu-devel] [RFC 04/13] migration: Create a snapshot thread to realize saving memory snapshot zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 05/13] migration: implement initialization work for snapshot zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 06/13] QEMUSizedBuffer: Introduce two help functions for qsb zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 07/13] savevm: Split qemu_savevm_state_complete_precopy() into two helper functions zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 08/13] snapshot: Save VM's device state into snapshot file zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 09/13] migration/postcopy-ram: fix some helper functions to support userfaultfd write-protect zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 10/13] snapshot: Enable the write-protect notification capability for VM's RAM zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 11/13] snapshot/migration: Save VM's RAM into snapshot file zhanghailiang
2016-01-07 12:20 ` [Qemu-devel] [RFC 12/13] migration/ram: Fix some helper functions' parameter to use PageSearchStatus zhanghailiang
2016-01-11 17:55   ` Dr. David Alan Gilbert
2016-01-12 12:59     ` Hailiang Zhang
2016-01-07 12:20 ` [Qemu-devel] [RFC 13/13] snapshot: Remove page's write-protect and copy the content during setup stage zhanghailiang
2016-07-13 17:52   ` Dr. David Alan Gilbert
2016-07-14  8:02     ` Hailiang Zhang
2016-07-04 12:22 ` [Qemu-devel] [RFC 00/13] Live memory snapshot based on userfaultfd Baptiste Reynal
2016-07-05  1:49   ` Hailiang Zhang
2016-07-05  9:57     ` Baptiste Reynal
2016-07-05 10:27       ` Hailiang Zhang
2016-08-18 15:56         ` Andrea Arcangeli
2016-08-20  6:31           ` Hailiang Zhang
2017-02-27 15:37             ` Christian Pinto
2017-02-28  1:48               ` Hailiang Zhang
2017-02-28  8:30                 ` Christian Pinto
2017-02-28 16:14                 ` Andrea Arcangeli
2017-03-01  1:08                   ` Hailiang Zhang
2017-03-09 11:34             ` [Qemu-devel] [RFC PATCH 0/4] ARM/ARM64 fixes for live " Christian Pinto
2017-03-09 11:34               ` [Qemu-devel] [RFC PATCH 1/4] migration/postcopy-ram: check pagefault flags in userfaultfd thread Christian Pinto
2017-03-09 11:34               ` Christian Pinto [this message]
2017-03-09 11:34               ` [Qemu-devel] [RFC PATCH 3/4] migration: snapshot thread Christian Pinto
2017-03-09 11:34               ` [Qemu-devel] [RFC PATCH 4/4] migration/postcopy-ram: ram_set_pages_wp fix Christian Pinto
2017-03-09 17:46               ` [Qemu-devel] [RFC PATCH 0/4] ARM/ARM64 fixes for live memory snapshot based on userfaultfd Dr. David Alan Gilbert
2017-03-10  8:15                 ` Christian Pinto
2016-09-06  3:39           ` [Qemu-devel] [RFC 00/13] Live " Hailiang Zhang
2016-09-18  2:14             ` Hailiang Zhang
2016-12-08 12:45               ` Hailiang Zhang
2016-07-05 14:59       ` Andrea Arcangeli
2016-07-13 18:02 ` Dr. David Alan Gilbert
2016-07-14 10:24   ` Hailiang Zhang
2016-07-14 11:43     ` Dr. David Alan Gilbert
2016-07-19  6:53       ` Hailiang Zhang

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=20170309113437.9667-3-c.pinto@virtualopensystems.com \
    --to=c.pinto@virtualopensystems.com \
    --cc=aarcange@redhat.com \
    --cc=amit.shah@redhat.com \
    --cc=b.reynal@virtualopensystems.com \
    --cc=dgilbert@redhat.com \
    --cc=hanweidong@huawei.com \
    --cc=peter.huangpeng@huawei.com \
    --cc=qemu-devel@nongnu.org \
    --cc=quintela@redhat.com \
    --cc=tech@virtualopensystems.com \
    --cc=zhang.zhanghailiang@huawei.com \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.