bdev_discard_supported: understand discard_granularity=0

Kernel documentation for the discard_granularity property says:

    A discard_granularity of 0 means that the device does not support
    discard functionality.

Some older kernels had drivers (notably loop, but also some USB-SATA
adapters) that would set the QUEUE_FLAG_DISCARD capability flag, but
have discard_granularity=0. Since 5.10 (torvalds/linux@b35fd7422c2f) the
discard entry point blkdev_issue_discard() has had a check for this,
which would immediately reject the call with EOPNOTSUPP, and throw a
scary diagnostic message into the log. See openzfs#16068.

Since 6.8, the block layer sets a non-zero default for
discard_granularity (torvalds/linux@3c407dc723bb), and a future kernel
will remove the check entirely[1].

As such, there's no good reason for us to enable discard when
discard_granularity=0. The kernel will never let the request go in
anyway; better that we just disable it so we can report it properly to
the user.

1. https://patchwork.kernel.org/project/linux-block/patch/[email protected]/

Sponsored-by: Klara, Inc.
Sponsored-by: Wasabi Technology, Inc.
Signed-off-by: Rob Norris <[email protected]>
(cherry picked from commit b181b2e)

include/os/linux/kernel/linux/blkdev_compat.h

@@ -563,9 +563,11 @@ static inline boolean_t
 bdev_discard_supported(struct block_device *bdev)
 {
 #if defined(HAVE_BDEV_MAX_DISCARD_SECTORS)
-	return (!!bdev_max_discard_sectors(bdev));
+	return (bdev_max_discard_sectors(bdev) > 0 &&
+	    bdev_discard_granularity(bdev) > 0);
 #elif defined(HAVE_BLK_QUEUE_DISCARD)
-	return (!!blk_queue_discard(bdev_get_queue(bdev)));
+	return (blk_queue_discard(bdev_get_queue(bdev)) > 0 &&
+	    bdev_get_queue(bdev)->limits.discard_granularity > 0);
 #else
 #error "Unsupported kernel"
 #endif