[qemu] hdparm tunable IDE write cache for HVM

qemu 0.8.2 has a flush callback to the storage backends, so now it is
possible to implement hdparm tunable IDE write cache enable/disable for
guest domains, allowing people to pick speed or data consistency on a
case by case basis.

As an added benefit, really large LBA48 IOs will now no longer be broken
up into smaller IOs on the host side.

From: Rik van Riel <riel@redhat.com>
Signed-off-by: Christian Limpach <Christian.Limpach@xensource.com>

diff --git a/tools/ioemu/block-bochs.c b/tools/ioemu/block-bochs.c
index d82c2392818d454778495efc2993a88c19d717ea..62317aff38a83ac6937de56baaa2155581151a09 100644 (file)
--- a/tools/ioemu/block-bochs.c
+++ b/tools/ioemu/block-bochs.c
@@ -91,7 +91,7 @@ static int bochs_open(BlockDriverState *bs, const char *filename)
     int fd, i;
     struct bochs_header bochs;
 
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE | O_SYNC);
+    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
     if (fd < 0) {
         fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
         if (fd < 0)

diff --git a/tools/ioemu/block-cloop.c b/tools/ioemu/block-cloop.c
index 0f56685664b8eaa08498ced94f9ce387ee61fa56..c617e1b64bc3d2b391515b1413d0edba5f040b36 100644 (file)
--- a/tools/ioemu/block-cloop.c
+++ b/tools/ioemu/block-cloop.c
@@ -55,7 +55,7 @@ static int cloop_open(BlockDriverState *bs, const char *filename)
     BDRVCloopState *s = bs->opaque;
     uint32_t offsets_size,max_compressed_block_size=1,i;
 
-    s->fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE | O_SYNC);
+    s->fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
     if (s->fd < 0)
         return -1;
     bs->read_only = 1;

diff --git a/tools/ioemu/block-cow.c b/tools/ioemu/block-cow.c
index ff73cac0f9a2d646be318702eccf03ea5e227446..6af8b749759910eca057e6dd430ae01d8f099086 100644 (file)
--- a/tools/ioemu/block-cow.c
+++ b/tools/ioemu/block-cow.c
@@ -69,7 +69,7 @@ static int cow_open(BlockDriverState *bs, const char *filename)
     struct cow_header_v2 cow_header;
     int64_t size;
 
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE | O_SYNC);
+    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
     if (fd < 0) {
         fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
         if (fd < 0)

diff --git a/tools/ioemu/block-qcow.c b/tools/ioemu/block-qcow.c
index d7fd9d534ba2b3b337a83942fd8d77ab12d66490..e5b52fb86108aab79620f5b541d781628f15af61 100644 (file)
--- a/tools/ioemu/block-qcow.c
+++ b/tools/ioemu/block-qcow.c
@@ -95,7 +95,7 @@ static int qcow_open(BlockDriverState *bs, const char *filename)
     int fd, len, i, shift;
     QCowHeader header;
     
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE | O_SYNC);
+    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
     if (fd < 0) {
         fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
         if (fd < 0)

diff --git a/tools/ioemu/block-vmdk.c b/tools/ioemu/block-vmdk.c
index 3143077f4948ae4e92683347beeac38622cc70d0..4cc3db84a1c4c8b9756316197567f1e4d318dcaf 100644 (file)
--- a/tools/ioemu/block-vmdk.c
+++ b/tools/ioemu/block-vmdk.c
@@ -96,7 +96,7 @@ static int vmdk_open(BlockDriverState *bs, const char *filename)
     uint32_t magic;
     int l1_size;
 
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE | O_SYNC);
+    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
     if (fd < 0) {
         fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
         if (fd < 0)

diff --git a/tools/ioemu/block.c b/tools/ioemu/block.c
index 210cb7a51ef12f69f4e7b20cdea570e7e3f77a33..00444abb19f34a7516e609d68cc367f285fc5d75 100644 (file)
--- a/tools/ioemu/block.c
+++ b/tools/ioemu/block.c
@@ -685,7 +685,7 @@ static int raw_open(BlockDriverState *bs, const char *filename)
     int rv;
 #endif
 
-    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE | O_SYNC);
+    fd = open(filename, O_RDWR | O_BINARY | O_LARGEFILE);
     if (fd < 0) {
         fd = open(filename, O_RDONLY | O_BINARY | O_LARGEFILE);
         if (fd < 0)

diff --git a/tools/ioemu/hw/ide.c b/tools/ioemu/hw/ide.c
index 92d33ef6e025b14570013f63fd49de3001dc25e4..8b070cc0cc9257691370b093e242b70ef9ea733c 100644 (file)
--- a/tools/ioemu/hw/ide.c
+++ b/tools/ioemu/hw/ide.c
@@ -305,6 +305,7 @@ typedef struct IDEState {
     PCIDevice *pci_dev;
     struct BMDMAState *bmdma;
     int drive_serial;
+    int write_cache;
     /* ide regs */
     uint8_t feature;
     uint8_t error;
@@ -789,6 +790,9 @@ static void ide_sector_write(IDEState *s)
     }
     ide_set_sector(s, sector_num + n);
     
+    if (!s->write_cache)
+        bdrv_flush(s->bs);
+    
 #ifdef TARGET_I386
     if (win2k_install_hack && ((++s->irq_count % 16) == 0)) {
         /* It seems there is a bug in the Windows 2000 installer HDD
@@ -863,6 +867,10 @@ static int ide_write_dma_cb(IDEState *s,
         transfer_size -= len;
         phys_addr += len;
     }
+    /* Ensure the data hit disk before telling the guest OS so. */
+    if (!s->write_cache)
+        bdrv_flush(s->bs);
+
     return transfer_size1 - transfer_size;
 }
 
@@ -1672,7 +1680,15 @@ static void ide_ioport_write(void *opaque, uint32_t addr, uint32_t val)
             /* XXX: valid for CDROM ? */
             switch(s->feature) {
             case 0x02: /* write cache enable */
+                s->write_cache = 1;
+                s->status = READY_STAT | SEEK_STAT;
+                ide_set_irq(s);
+                break;
             case 0x82: /* write cache disable */
+                s->write_cache = 0;
+                s->status = READY_STAT | SEEK_STAT;
+                ide_set_irq(s);
+                break;
             case 0xaa: /* read look-ahead enable */
             case 0x55: /* read look-ahead disable */
                 s->status = READY_STAT | SEEK_STAT;
@@ -2090,6 +2106,7 @@ static void ide_init2(IDEState *ide_state,
         s->irq = irq;
         s->sector_write_timer = qemu_new_timer(vm_clock, 
                                                ide_sector_write_timer_cb, s);
+        s->write_cache = 0;
         ide_reset(s);
     }
 }