block: Cater to iscsi with non-power-of-2 discard

Dell Equallogic iSCSI SANs have a very unusual advertised geometry:

$ iscsi-inq -e 1 -c $((0xb0)) iscsi://XXX/0
wsnz:0
maximum compare and write length:1
optimal transfer length granularity:0
maximum transfer length:0
optimal transfer length:0
maximum prefetch xdread xdwrite transfer length:0
maximum unmap lba count:30720
maximum unmap block descriptor count:2
optimal unmap granularity:30720
ugavalid:1
unmap granularity alignment:0
maximum write same length:30720

which says that both the maximum and the optimal discard size
is 15M.  It is not immediately apparent if the device allows
discard requests not aligned to the optimal size, nor if it
allows discards at a finer granularity than the optimal size.

I tried to find details in the SCSI Commands Reference Manual
Rev. A on what valid values of maximum and optimal sizes are
permitted, but while that document mentions a "Block Limits
VPD Page", I couldn't actually find documentation of that page
or what values it would have, or if a SCSI device has an
advertisement of its minimal unmap granularity.  So it is not
obvious to me whether the Dell Equallogic device is compliance
with the SCSI specification.

Fortunately, it is easy enough to support non-power-of-2 sizing,
even if it means we are less efficient than truly possible when
targetting that device (for example, it means that we refuse to
unmap anything that is not a multiple of 15M and aligned to a
15M boundary, even if the device truly does support a smaller
granularity where unmapping actually works).

Reported-by: Peter Lieven <[email protected]>
Signed-off-by: Eric Blake <[email protected]>
Message-Id: <1469129688[email protected]>
Acked-by: Stefan Hajnoczi <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>

diff --git a/block/io.c b/block/io.c
index 7323f0fb7b63bb967f89488fb08965a8fe5cb61f..d5493ba349cdeaf96aafed85e4160157462394d2 100644 (file)
--- a/block/io.c
+++ b/block/io.c
@@ -1180,10 +1180,11 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs,
     int alignment = MAX(bs->bl.pwrite_zeroes_alignment,
                         bs->bl.request_alignment);
 
-    assert(is_power_of_2(alignment));
-    head = offset & (alignment - 1);
-    tail = (offset + count) & (alignment - 1);
-    max_write_zeroes &= ~(alignment - 1);
+    assert(alignment % bs->bl.request_alignment == 0);
+    head = offset % alignment;
+    tail = (offset + count) % alignment;
+    max_write_zeroes = QEMU_ALIGN_DOWN(max_write_zeroes, alignment);
+    assert(max_write_zeroes >= bs->bl.request_alignment);
 
     while (count > 0 && !ret) {
         int num = count;
@@ -2429,9 +2430,10 @@ int coroutine_fn bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset,
 
     /* Discard is advisory, so ignore any unaligned head or tail */
     align = MAX(bs->bl.pdiscard_alignment, bs->bl.request_alignment);
-    assert(is_power_of_2(align));
-    head = MIN(count, -offset & (align - 1));
+    assert(align % bs->bl.request_alignment == 0);
+    head = offset % align;
     if (head) {
+        head = MIN(count, align - head);
         count -= head;
         offset += head;
     }
@@ -2449,6 +2451,7 @@ int coroutine_fn bdrv_co_pdiscard(BlockDriverState *bs, int64_t offset,
 
     max_pdiscard = QEMU_ALIGN_DOWN(MIN_NON_ZERO(bs->bl.max_pdiscard, INT_MAX),
                                    align);
+    assert(max_pdiscard);
 
     while (count > 0) {
         int ret;

diff --git a/include/block/block_int.h b/include/block/block_int.h
index 1fe0fd9ff61b1dbbfce633293b0178fab10f6de0..47665be81e2392810a3bc8eff79b7747302a663a 100644 (file)
--- a/include/block/block_int.h
+++ b/include/block/block_int.h
@@ -330,36 +330,39 @@ typedef struct BlockLimits {
      * otherwise. */
     uint32_t request_alignment;
 
-    /* maximum number of bytes that can be discarded at once (since it
-     * is signed, it must be < 2G, if set), should be multiple of
+    /* Maximum number of bytes that can be discarded at once (since it
+     * is signed, it must be < 2G, if set). Must be multiple of
      * pdiscard_alignment, but need not be power of 2. May be 0 if no
      * inherent 32-bit limit */
     int32_t max_pdiscard;
 
-    /* optimal alignment for discard requests in bytes, must be power
-     * of 2, less than max_pdiscard if that is set, and multiple of
-     * bl.request_alignment. May be 0 if bl.request_alignment is good
-     * enough */
+    /* Optimal alignment for discard requests in bytes. A power of 2
+     * is best but not mandatory.  Must be a multiple of
+     * bl.request_alignment, and must be less than max_pdiscard if
+     * that is set. May be 0 if bl.request_alignment is good enough */
     uint32_t pdiscard_alignment;
 
-    /* maximum number of bytes that can zeroized at once (since it is
-     * signed, it must be < 2G, if set), should be multiple of
+    /* Maximum number of bytes that can zeroized at once (since it is
+     * signed, it must be < 2G, if set). Must be multiple of
      * pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */
     int32_t max_pwrite_zeroes;
 
-    /* optimal alignment for write zeroes requests in bytes, must be
-     * power of 2, less than max_pwrite_zeroes if that is set, and
-     * multiple of bl.request_alignment. May be 0 if
-     * bl.request_alignment is good enough */
+    /* Optimal alignment for write zeroes requests in bytes. A power
+     * of 2 is best but not mandatory.  Must be a multiple of
+     * bl.request_alignment, and must be less than max_pwrite_zeroes
+     * if that is set. May be 0 if bl.request_alignment is good
+     * enough */
     uint32_t pwrite_zeroes_alignment;
 
-    /* optimal transfer length in bytes (must be power of 2, and
-     * multiple of bl.request_alignment), or 0 if no preferred size */
+    /* Optimal transfer length in bytes.  A power of 2 is best but not
+     * mandatory.  Must be a multiple of bl.request_alignment, or 0 if
+     * no preferred size */
     uint32_t opt_transfer;
 
-    /* maximal transfer length in bytes (need not be power of 2, but
-     * should be multiple of opt_transfer), or 0 for no 32-bit limit.
-     * For now, anything larger than INT_MAX is clamped down. */
+    /* Maximal transfer length in bytes.  Need not be power of 2, but
+     * must be multiple of opt_transfer and bl.request_alignment, or 0
+     * for no 32-bit limit.  For now, anything larger than INT_MAX is
+     * clamped down. */
     uint32_t max_transfer;
 
     /* memory alignment, in bytes so that no bounce buffer is needed */