Skip to contentSkip to conversation

tools: build fix for blktap2

PATCH In review Christoph Egger 2 participants 3 messages Feb 4, 2010

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tools: build fix for blktap2

Hi!

Attached patch makes blktap2 build again.

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Andrew Bowd, Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632
tools/blktap2/drivers/tapdisk-queue.c +6 -0
diff -r e15391c2e6f1 tools/blktap2/drivers/tapdisk-queue.c
@@ -284,11 +284,14 @@ struct lio {
 
 #define LIO_FLAG_EVENTFD        (1<<0)
 
+#if defined(__linux__)
+/* XXX abstraction violation: move this into a linux specific file */
 static int
 tapdisk_lio_check_resfd(void)
 {
 	return tapdisk_linux_version() >= KERNEL_VERSION(2, 6, 22);
 }
+#endif
 
 static void
 tapdisk_lio_destroy_aio(struct tqueue *queue)
@@ -372,8 +375,11 @@ tapdisk_lio_setup_aio(struct tqueue *que
 	 * if not, fall back to the poll fd patch.
 	 */
 
+#if defined(__linux__)
+/* XXX abstraction violation: move this into a linux specific file */
 	err = !tapdisk_lio_check_resfd();
 	if (!err)
+#endif
 		err = __lio_setup_aio_eventfd(queue, qlen);
 	if (err)
 		err = __lio_setup_aio_poll(queue, qlen);

_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xensource.com/xen-devel

#1 lore raw
Quoted 2 lines · Christoph Egger

Hi!

Hi!

Quoted 6 lines

Attached patch makes blktap2 build again.

#define LIO_FLAG_EVENTFD (1<<0)

tools/blktap2/drivers/tapdisk-queue.c struct lio { · line 294

Quoted 8 lines
tools/blktap2/drivers/tapdisk-queue.c +3 -0
+#if defined(__linux__)
+/* XXX abstraction violation: move this into a linux specific file */
 static int
 tapdisk_lio_check_resfd(void)
 {
        return tapdisk_linux_version() >= KERNEL_VERSION(2, 6, 22);
 }
+#endif

Ah right, I forgot to mask out KERNEL_VERSION itself.

tapdisk_linux_version is defined however, turning into an ENOSYS.

How about just running and failing that check?

tools/blktap2/drivers/tapdisk-queue.c tapdisk_lio_setup_aio(struct tqueue *que · line 383

Quoted 12 lines

static void
tapdisk_lio_destroy_aio(struct tqueue *queue)

tools/blktap2/drivers/tapdisk-queue.c +3 -0
@@ -372,8 +375,11 @@ tapdisk_lio_setup_aio(struct tqueue *que
         * if not, fall back to the poll fd patch.
         */
 
+#if defined(__linux__)
+/* XXX abstraction violation: move this into a linux specific file */
        err = !tapdisk_lio_check_resfd();
        if (!err)
+#endif
                err = __lio_setup_aio_eventfd(queue, qlen);

I don't think you wanted this, unless you're sure what you keep calling.

tools/blktap2/drivers/tapdisk-queue.c tapdisk_lio_setup_aio(struct tqueue *que · line 385

Quoted 2 lines
tools/blktap2/drivers/tapdisk-queue.c +0 -0
        if (err)
                err = __lio_setup_aio_poll(queue, qlen);

Just to understand the FreeBSD situation a little better -- it actually does come with a mechanism equivalent to the poll patch?

Cheers,
Daniel

#2 lore raw
Quoted 22 lines · On Thursday 04 February 2010 13:39:44 Daniel Stodden wrote:
Quoted 1 line · Christoph Egger

Hi!

Hi!

Quoted 6 lines

Attached patch makes blktap2 build again.

#define LIO_FLAG_EVENTFD (1<<0)

tools/blktap2/drivers/tapdisk-queue.c struct lio { · line 294

Quoted 8 lines
tools/blktap2/drivers/tapdisk-queue.c +3 -0
+#if defined(__linux__)
+/* XXX abstraction violation: move this into a linux specific file */
 static int
 tapdisk_lio_check_resfd(void)
 {
        return tapdisk_linux_version() >= KERNEL_VERSION(2, 6, 22);
 }
+#endif

Ah right, I forgot to mask out KERNEL_VERSION itself.

Quoted 3 lines

tapdisk_linux_version is defined however, turning into an ENOSYS.

How about just running and failing that check?

Depends on what method should be used.

Quoted 7 lines
Quoted 7 lines

static void
tapdisk_lio_destroy_aio(struct tqueue *queue)

tools/blktap2/drivers/tapdisk-queue.c +0 -0
@@ -372,8 +375,11 @@ tapdisk_lio_setup_aio(struct tqueue *que
         * if not, fall back to the poll fd patch.
         */
 
Quoted 8 lines

tools/blktap2/drivers/tapdisk-queue.c tapdisk_lio_setup_aio(struct tqueue *que · line 383

Quoted 6 lines
tools/blktap2/drivers/tapdisk-queue.c +3 -0
+#if defined(__linux__)
+/* XXX abstraction violation: move this into a linux specific file */
        err = !tapdisk_lio_check_resfd();
        if (!err)
+#endif
                err = __lio_setup_aio_eventfd(queue, qlen);

I don't think you wanted this, unless you're sure what you keep calling.

Is the Linux event comparable to this?
http://netbsd.gw.com/cgi-bin/man-cgi?event++NetBSD-current

Quoted 5 lines

tools/blktap2/drivers/tapdisk-queue.c tapdisk_lio_setup_aio(struct tqueue *que · line 385

Quoted 2 lines
tools/blktap2/drivers/tapdisk-queue.c +0 -0
        if (err)
                err = __lio_setup_aio_poll(queue, qlen);

Just to understand the FreeBSD situation a little better -- it actually does come with a mechanism equivalent to the poll patch?

NetBSD supports POSIX AIO and has poll(2).
http://netbsd.gw.com/cgi-bin/man-cgi?poll++NetBSD-current

Christoph

Signature · 6 lines
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Karl-Hammerschmidt-Str. 34, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Andrew Bowd, Thomas M. McCoy, Giuliano Meroni
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632