File kvm-qemu-preXX-scsi-disk-lazily-allocate-bounce-buffer.patch of Package kvm

From 7285477ab11831b1cf56e45878a89170dd06d9b9 Mon Sep 17 00:00:00 2001
From: Paolo Bonzini <pbonzini@redhat.com>
Date: Fri, 16 Sep 2011 16:40:04 +0200
Subject: [PATCH] scsi-disk: lazily allocate bounce buffer

It will not be needed for reads and writes if the HBA provides a sglist.
In addition, this lets scsi-disk refuse commands with an excessive
allocation length, as well as limit memory on usual well-behaved guests.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
 hw/scsi-disk.c |   44 +++++++++++++++++++++++++++++++++-----------
 1 files changed, 33 insertions(+), 11 deletions(-)

Index: qemu-kvm-0.15.1/hw/scsi-disk.c
===================================================================
--- qemu-kvm-0.15.1.orig/hw/scsi-disk.c
+++ qemu-kvm-0.15.1/hw/scsi-disk.c
@@ -54,6 +54,7 @@ typedef struct SCSIDiskReq {
     /* Both sector and sector_count are in terms of qemu 512 byte blocks.  */
     uint64_t sector;
     uint32_t sector_count;
+    uint32_t buflen;
     struct iovec iov;
     QEMUIOVector qiov;
     uint32_t status;
@@ -78,18 +79,15 @@ struct SCSIDiskState
 };
 
 static int scsi_handle_rw_error(SCSIDiskReq *r, int error, int type);
-static int scsi_disk_emulate_command(SCSIDiskReq *r, uint8_t *outbuf);
+static int scsi_disk_emulate_command(SCSIDiskReq *r);
 
 static SCSIRequest *scsi_new_request(SCSIDevice *d, uint32_t tag,
                                      uint32_t lun, void *hba_private)
 {
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, d);
     SCSIRequest *req;
-    SCSIDiskReq *r;
 
     req = scsi_req_alloc(sizeof(SCSIDiskReq), &s->qdev, tag, lun, hba_private);
-    r = DO_UPCAST(SCSIDiskReq, req, req);
-    r->iov.iov_base = qemu_blockalign(s->bs, SCSI_DMA_BUF_SIZE);
     return req;
 }
 
@@ -97,7 +95,9 @@ static void scsi_free_request(SCSIReques
 {
     SCSIDiskReq *r = DO_UPCAST(SCSIDiskReq, req, req);
 
-    qemu_vfree(r->iov.iov_base);
+    if (r->iov.iov_base) {
+        qemu_vfree(r->iov.iov_base);
+    }
 }
 
 static void scsi_disk_clear_sense(SCSIDiskState *s)
@@ -136,7 +136,13 @@ static void scsi_cancel_io(SCSIRequest *
 
 static uint32_t scsi_init_iovec(SCSIDiskReq *r)
 {
-    r->iov.iov_len = MIN(r->sector_count * 512, SCSI_DMA_BUF_SIZE);
+    SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, r->req.dev);
+
+    if (!r->iov.iov_base) {
+        r->buflen = SCSI_DMA_BUF_SIZE;
+        r->iov.iov_base = qemu_blockalign(s->bs, r->buflen);
+    }
+    r->iov.iov_len = MIN(r->sector_count * 512, r->buflen);
     qemu_iovec_init_external(&r->qiov, &r->iov, 1);
     return r->qiov.size / 512;
 }
@@ -320,7 +326,7 @@ static void scsi_dma_restart_bh(void *op
                 scsi_write_data(&r->req);
                 break;
             case SCSI_REQ_STATUS_RETRY_FLUSH:
-                ret = scsi_disk_emulate_command(r, r->iov.iov_base);
+                ret = scsi_disk_emulate_command(r);
                 if (ret == 0) {
                     scsi_command_complete(r, GOOD, SENSE_CODE(NO_SENSE));
                 }
@@ -820,14 +826,32 @@ static int scsi_disk_emulate_read_toc(SC
     return toclen;
 }
 
-static int scsi_disk_emulate_command(SCSIDiskReq *r, uint8_t *outbuf)
+static int scsi_disk_emulate_command(SCSIDiskReq *r)
 {
     SCSIRequest *req = &r->req;
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, req->dev);
     uint64_t nb_sectors;
+    uint8_t *outbuf;
     int buflen = 0;
     int ret;
 
+    if (!r->iov.iov_base) {
+        /*
+         * FIXME: we shouldn't return anything bigger than 4k, but the code
+         * requires the buffer to be as big as req->cmd.xfer in several
+         * places.  So, do not allow CDBs with a very large ALLOCATION
+         * LENGTH.  The real fix would be to modify scsi_read_data and
+         * dma_buf_read, so that they return data beyond the buflen
+         * as all zeros.
+         */
+        if (req->cmd.xfer > 65536) {
+            goto illegal_request;
+        }
+        r->buflen = MAX(4096, req->cmd.xfer);
+        r->iov.iov_base = qemu_blockalign(s->bs, r->buflen);
+    }
+
+    outbuf = r->iov.iov_base;
     switch (req->cmd.buf[0]) {
     case TEST_UNIT_READY:
         if (!bdrv_is_inserted(s->bs))
@@ -1005,11 +1029,9 @@ static int32_t scsi_send_command(SCSIReq
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, req->dev);
     int32_t len;
     uint8_t command;
-    uint8_t *outbuf;
     int rc;
 
     command = buf[0];
-    outbuf = (uint8_t *)r->iov.iov_base;
     DPRINTF("Command: lun=%d tag=0x%x data=0x%02x", req->lun, req->tag, buf[0]);
 
     if (scsi_req_parse(&r->req, buf) != 0) {
@@ -1056,7 +1078,7 @@ static int32_t scsi_send_command(SCSIReq
     case REPORT_LUNS:
     case VERIFY:
     case REZERO_UNIT:
-        rc = scsi_disk_emulate_command(r, outbuf);
+        rc = scsi_disk_emulate_command(r);
         if (rc < 0) {
             return 0;
         }