958e1b
From 3555c9b0e05c410028022a1b1af34ad7a2e83e98 Mon Sep 17 00:00:00 2001
958e1b
From: Laszlo Ersek <lersek@redhat.com>
958e1b
Date: Fri, 7 Nov 2014 17:17:59 +0100
958e1b
Subject: [PATCH 12/41] dump: add API to write dump_bitmap
958e1b
958e1b
Message-id: <1415380693-16593-13-git-send-email-lersek@redhat.com>
958e1b
Patchwork-id: 62198
958e1b
O-Subject: [RHEL-7.1 qemu-kvm PATCH 12/26] dump: add API to write dump_bitmap
958e1b
Bugzilla: 1157798
958e1b
RH-Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
958e1b
RH-Acked-by: Luiz Capitulino <lcapitulino@redhat.com>
958e1b
RH-Acked-by: dgibson <dgibson@redhat.com>
958e1b
958e1b
From: qiaonuohan <qiaonuohan@cn.fujitsu.com>
958e1b
958e1b
functions are used to write 1st and 2nd dump_bitmap of kdump-compressed format,
958e1b
which is used to indicate whether the corresponded page is existed in vmcore.
958e1b
1st and 2nd dump_bitmap are same, because dump level is specified to 1 here.
958e1b
958e1b
Signed-off-by: Qiao Nuohan <qiaonuohan@cn.fujitsu.com>
958e1b
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
958e1b
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
958e1b
(cherry picked from commit d0686c7291fe8f0210e7a666f80892fa71395510)
958e1b
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
958e1b
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
958e1b
---
958e1b
 dump.c                | 164 ++++++++++++++++++++++++++++++++++++++++++++++++++
958e1b
 include/sysemu/dump.h |   2 +
958e1b
 2 files changed, 166 insertions(+)
958e1b
958e1b
diff --git a/dump.c b/dump.c
958e1b
index 4d135fd..f416093 100644
958e1b
--- a/dump.c
958e1b
+++ b/dump.c
958e1b
@@ -1007,6 +1007,170 @@ static int write_dump_header(DumpState *s)
958e1b
     }
958e1b
 }
958e1b
 
958e1b
+/*
958e1b
+ * set dump_bitmap sequencely. the bit before last_pfn is not allowed to be
958e1b
+ * rewritten, so if need to set the first bit, set last_pfn and pfn to 0.
958e1b
+ * set_dump_bitmap will always leave the recently set bit un-sync. And setting
958e1b
+ * (last bit + sizeof(buf) * 8) to 0 will do flushing the content in buf into
958e1b
+ * vmcore, ie. synchronizing un-sync bit into vmcore.
958e1b
+ */
958e1b
+static int set_dump_bitmap(uint64_t last_pfn, uint64_t pfn, bool value,
958e1b
+                           uint8_t *buf, DumpState *s)
958e1b
+{
958e1b
+    off_t old_offset, new_offset;
958e1b
+    off_t offset_bitmap1, offset_bitmap2;
958e1b
+    uint32_t byte, bit;
958e1b
+
958e1b
+    /* should not set the previous place */
958e1b
+    assert(last_pfn <= pfn);
958e1b
+
958e1b
+    /*
958e1b
+     * if the bit needed to be set is not cached in buf, flush the data in buf
958e1b
+     * to vmcore firstly.
958e1b
+     * making new_offset be bigger than old_offset can also sync remained data
958e1b
+     * into vmcore.
958e1b
+     */
958e1b
+    old_offset = BUFSIZE_BITMAP * (last_pfn / PFN_BUFBITMAP);
958e1b
+    new_offset = BUFSIZE_BITMAP * (pfn / PFN_BUFBITMAP);
958e1b
+
958e1b
+    while (old_offset < new_offset) {
958e1b
+        /* calculate the offset and write dump_bitmap */
958e1b
+        offset_bitmap1 = s->offset_dump_bitmap + old_offset;
958e1b
+        if (write_buffer(s->fd, offset_bitmap1, buf,
958e1b
+                         BUFSIZE_BITMAP) < 0) {
958e1b
+            return -1;
958e1b
+        }
958e1b
+
958e1b
+        /* dump level 1 is chosen, so 1st and 2nd bitmap are same */
958e1b
+        offset_bitmap2 = s->offset_dump_bitmap + s->len_dump_bitmap +
958e1b
+                         old_offset;
958e1b
+        if (write_buffer(s->fd, offset_bitmap2, buf,
958e1b
+                         BUFSIZE_BITMAP) < 0) {
958e1b
+            return -1;
958e1b
+        }
958e1b
+
958e1b
+        memset(buf, 0, BUFSIZE_BITMAP);
958e1b
+        old_offset += BUFSIZE_BITMAP;
958e1b
+    }
958e1b
+
958e1b
+    /* get the exact place of the bit in the buf, and set it */
958e1b
+    byte = (pfn % PFN_BUFBITMAP) / CHAR_BIT;
958e1b
+    bit = (pfn % PFN_BUFBITMAP) % CHAR_BIT;
958e1b
+    if (value) {
958e1b
+        buf[byte] |= 1u << bit;
958e1b
+    } else {
958e1b
+        buf[byte] &= ~(1u << bit);
958e1b
+    }
958e1b
+
958e1b
+    return 0;
958e1b
+}
958e1b
+
958e1b
+/*
958e1b
+ * exam every page and return the page frame number and the address of the page.
958e1b
+ * bufptr can be NULL. note: the blocks here is supposed to reflect guest-phys
958e1b
+ * blocks, so block->target_start and block->target_end should be interal
958e1b
+ * multiples of the target page size.
958e1b
+ */
958e1b
+static bool get_next_page(GuestPhysBlock **blockptr, uint64_t *pfnptr,
958e1b
+                          uint8_t **bufptr, DumpState *s)
958e1b
+{
958e1b
+    GuestPhysBlock *block = *blockptr;
958e1b
+    hwaddr addr;
958e1b
+    uint8_t *buf;
958e1b
+
958e1b
+    /* block == NULL means the start of the iteration */
958e1b
+    if (!block) {
958e1b
+        block = QTAILQ_FIRST(&s->guest_phys_blocks.head);
958e1b
+        *blockptr = block;
958e1b
+        assert(block->target_start % s->page_size == 0);
958e1b
+        assert(block->target_end % s->page_size == 0);
958e1b
+        *pfnptr = paddr_to_pfn(block->target_start, s->page_shift);
958e1b
+        if (bufptr) {
958e1b
+            *bufptr = block->host_addr;
958e1b
+        }
958e1b
+        return true;
958e1b
+    }
958e1b
+
958e1b
+    *pfnptr = *pfnptr + 1;
958e1b
+    addr = pfn_to_paddr(*pfnptr, s->page_shift);
958e1b
+
958e1b
+    if ((addr >= block->target_start) &&
958e1b
+        (addr + s->page_size <= block->target_end)) {
958e1b
+        buf = block->host_addr + (addr - block->target_start);
958e1b
+    } else {
958e1b
+        /* the next page is in the next block */
958e1b
+        block = QTAILQ_NEXT(block, next);
958e1b
+        *blockptr = block;
958e1b
+        if (!block) {
958e1b
+            return false;
958e1b
+        }
958e1b
+        assert(block->target_start % s->page_size == 0);
958e1b
+        assert(block->target_end % s->page_size == 0);
958e1b
+        *pfnptr = paddr_to_pfn(block->target_start, s->page_shift);
958e1b
+        buf = block->host_addr;
958e1b
+    }
958e1b
+
958e1b
+    if (bufptr) {
958e1b
+        *bufptr = buf;
958e1b
+    }
958e1b
+
958e1b
+    return true;
958e1b
+}
958e1b
+
958e1b
+static int write_dump_bitmap(DumpState *s)
958e1b
+{
958e1b
+    int ret = 0;
958e1b
+    uint64_t last_pfn, pfn;
958e1b
+    void *dump_bitmap_buf;
958e1b
+    size_t num_dumpable;
958e1b
+    GuestPhysBlock *block_iter = NULL;
958e1b
+
958e1b
+    /* dump_bitmap_buf is used to store dump_bitmap temporarily */
958e1b
+    dump_bitmap_buf = g_malloc0(BUFSIZE_BITMAP);
958e1b
+
958e1b
+    num_dumpable = 0;
958e1b
+    last_pfn = 0;
958e1b
+
958e1b
+    /*
958e1b
+     * exam memory page by page, and set the bit in dump_bitmap corresponded
958e1b
+     * to the existing page.
958e1b
+     */
958e1b
+    while (get_next_page(&block_iter, &pfn, NULL, s)) {
958e1b
+        ret = set_dump_bitmap(last_pfn, pfn, true, dump_bitmap_buf, s);
958e1b
+        if (ret < 0) {
958e1b
+            dump_error(s, "dump: failed to set dump_bitmap.\n");
958e1b
+            ret = -1;
958e1b
+            goto out;
958e1b
+        }
958e1b
+
958e1b
+        last_pfn = pfn;
958e1b
+        num_dumpable++;
958e1b
+    }
958e1b
+
958e1b
+    /*
958e1b
+     * set_dump_bitmap will always leave the recently set bit un-sync. Here we
958e1b
+     * set last_pfn + PFN_BUFBITMAP to 0 and those set but un-sync bit will be
958e1b
+     * synchronized into vmcore.
958e1b
+     */
958e1b
+    if (num_dumpable > 0) {
958e1b
+        ret = set_dump_bitmap(last_pfn, last_pfn + PFN_BUFBITMAP, false,
958e1b
+                              dump_bitmap_buf, s);
958e1b
+        if (ret < 0) {
958e1b
+            dump_error(s, "dump: failed to sync dump_bitmap.\n");
958e1b
+            ret = -1;
958e1b
+            goto out;
958e1b
+        }
958e1b
+    }
958e1b
+
958e1b
+    /* number of dumpable pages that will be dumped later */
958e1b
+    s->num_dumpable = num_dumpable;
958e1b
+
958e1b
+out:
958e1b
+    g_free(dump_bitmap_buf);
958e1b
+
958e1b
+    return ret;
958e1b
+}
958e1b
+
958e1b
 static ram_addr_t get_start_block(DumpState *s)
958e1b
 {
958e1b
     GuestPhysBlock *block;
958e1b
diff --git a/include/sysemu/dump.h b/include/sysemu/dump.h
958e1b
index dfee238..6d4d0bc 100644
958e1b
--- a/include/sysemu/dump.h
958e1b
+++ b/include/sysemu/dump.h
958e1b
@@ -39,6 +39,8 @@
958e1b
 #define PHYS_BASE                   (0)
958e1b
 #define DUMP_LEVEL                  (1)
958e1b
 #define DISKDUMP_HEADER_BLOCKS      (1)
958e1b
+#define BUFSIZE_BITMAP              (TARGET_PAGE_SIZE)
958e1b
+#define PFN_BUFBITMAP               (CHAR_BIT * BUFSIZE_BITMAP)
958e1b
 
958e1b
 typedef struct ArchDumpInfo {
958e1b
     int d_machine;  /* Architecture */
958e1b
-- 
958e1b
1.8.3.1
958e1b