[PATCH 0/2] tools/virtio: fix the build and the posix_memalign() checks

[PATCH 0/2] tools/virtio: fix the build and the posix_memalign() checks

[Liam R. Howlett (Google)]

Working on other tools, I came across the fact that posix_memalign()'s
return is not the standard negative error.  Scanning the git repo, I
came across the virtio issue with the return.  Testing turned out the be
broke as well, so there's a patch to fix testing prior to fixing the
caller.

Liam R. Howlett (Google) (2):
  tools/virtio: add a shim for linux/device-id/virtio.h
  tools/virtio: check posix_memalign() for failure

 tools/virtio/linux/device-id/virtio.h | 2 ++
 tools/virtio/linux/kernel.h           | 3 ++-
 tools/virtio/vhost_net_test.c         | 2 +-
 tools/virtio/virtio_test.c            | 2 +-
 4 files changed, 6 insertions(+), 3 deletions(-)
 create mode 100644 tools/virtio/linux/device-id/virtio.h

-- 
2.47.3

[PATCH 1/2] tools/virtio: add a shim for linux/device-id/virtio.h

[Liam R. Howlett (Google)]

include/linux/virtio.h takes struct virtio_device_id from
<linux/device-id/virtio.h>, which the tools build cannot find.

Add a shim for it beside the others, pointing at the real header.

Fixes: ecca1d63c1ea ("Replace <linux/mod_devicetable.h> by more specific <linux/device-id/*.h> (headers)")
Signed-off-by: Liam R. Howlett (Google) <liam@infradead.org>
---
 tools/virtio/linux/device-id/virtio.h | 2 ++
 1 file changed, 2 insertions(+)
 create mode 100644 tools/virtio/linux/device-id/virtio.h

diff --git a/tools/virtio/linux/device-id/virtio.h b/tools/virtio/linux/device-id/virtio.h
new file mode 100644
index 0000000000000..3c3e8c514c8d4
--- /dev/null
+++ b/tools/virtio/linux/device-id/virtio.h
@@ -0,0 +1,2 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#include <../../include/linux/device-id/virtio.h>
-- 
2.47.3

[PATCH 2/2] tools/virtio: check posix_memalign() for failure

[Liam R. Howlett (Google)]

posix_memalign() returns zero or a positive error number and does not set
errno.  virtio_test and vhost_net_test assert that the return is not
negative, which will always be true, and the userspace __get_free_page()
ignores it, so a failed allocation continues without setting the pointer

Assert a zero return in the tests, and return 0 from __get_free_page() on
failure, as the kernel's does.

Signed-off-by: Liam R. Howlett (Google) <liam@infradead.org>
---
 tools/virtio/linux/kernel.h   | 3 ++-
 tools/virtio/vhost_net_test.c | 2 +-
 tools/virtio/virtio_test.c    | 2 +-
 3 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/virtio/linux/kernel.h b/tools/virtio/linux/kernel.h
index 104abf9d1aee5..3497660c690a9 100644
--- a/tools/virtio/linux/kernel.h
+++ b/tools/virtio/linux/kernel.h
@@ -106,7 +106,8 @@ static inline unsigned long __get_free_page(gfp_t gfp)
 {
 	void *p;
 
-	posix_memalign(&p, PAGE_SIZE, PAGE_SIZE);
+	if (posix_memalign(&p, PAGE_SIZE, PAGE_SIZE))
+		return 0;
 	return (unsigned long)p;
 }
 
diff --git a/tools/virtio/vhost_net_test.c b/tools/virtio/vhost_net_test.c
index 566e15420bb65..3122ea0970568 100644
--- a/tools/virtio/vhost_net_test.c
+++ b/tools/virtio/vhost_net_test.c
@@ -197,7 +197,7 @@ static void vq_info_add(struct vdev_info *dev, int idx, int num, int fd)
 	info->idx = idx;
 	info->kick = eventfd(0, EFD_NONBLOCK);
 	r = posix_memalign(&info->ring, 4096, vring_size(num, 4096));
-	assert(r >= 0);
+	assert(!r);
 	vq_reset(info, num, &dev->vdev);
 	vhost_vq_setup(dev, info);
 
diff --git a/tools/virtio/virtio_test.c b/tools/virtio/virtio_test.c
index 028f54e6854a8..9d61463f6e09b 100644
--- a/tools/virtio/virtio_test.c
+++ b/tools/virtio/virtio_test.c
@@ -116,7 +116,7 @@ static void vq_info_add(struct vdev_info *dev, int num)
 	info->kick = eventfd(0, EFD_NONBLOCK);
 	info->call = eventfd(0, EFD_NONBLOCK);
 	r = posix_memalign(&info->ring, 4096, vring_size(num, 4096));
-	assert(r >= 0);
+	assert(!r);
 	vq_reset(info, num, &dev->vdev);
 	vhost_vq_setup(dev, info);
 	dev->fds[info->idx].fd = info->call;
-- 
2.47.3