bhyve/virtio-scsi: Don't invoke iov_to_buf() in an assert() expression

[rosenfeld_grumpf.hope-2000.org]

Summary

If anyone would build bhyve with -DNDEBUG, any code in the expression
in assert() won't be executed. Remove the assertion and check the
returned size from iov_to_buf() explicitly, exiting if it is wrong.

[emaste]

usr.sbin/bhyve/pci_virtio_scsi.c

754

[rosenfeld_grumpf.hope-2000.org]

Review from emaste.

[des]

usr.sbin/bhyve/pci_virtio_scsi.c

752

This still fails to build with gcc.

[rosenfeld_grumpf.hope-2000.org]

usr.sbin/bhyve/pci_virtio_scsi.c

752

That's not surprising, given this is a fix for a different issue. Please see https://reviews.freebsd.org/D55800 for the proper fix of the build failure with gcc.

[rosenfeld_grumpf.hope-2000.org]

Rebase.

[emaste]

Or, something along the lines of:

size_t res __maybe_unused;
res = iov_to_buf(...);
assert (res == VTSCSI_IN_HEADER_LEN(q->vsq_sc);

[rosenfeld_grumpf.hope-2000.org]

In D55803#1288531, @emaste wrote:

Or, something along the lines of:

size_t res __maybe_unused;
res = iov_to_buf(...);
assert (res == VTSCSI_IN_HEADER_LEN(q->vsq_sc);

There doesn't seem to be a general definition of __maybe_unused in FreeBSD, or I'm just too dense to find it. Some externals sources like ZFS seem to have it, though.

Perhaps I could use maybe_unused, but that's a C23 feature.

[markj]

The commit should include a Fixes: 2a514d377b37 ("bhyve/virtio-scsi: Preallocate all I/O requests") tag.

[emaste]

There doesn't seem to be a general definition of __maybe_unused in FreeBSD

Ah, yes. We seem two have three copies of the #define in our tree, two under sys/, but it's not generally available. We could perhaps define it in cdefs.h, but independent of this change.

[rosenfeld_grumpf.hope-2000.org]

In D55803#1291246, @emaste wrote:

There doesn't seem to be a general definition of __maybe_unused in FreeBSD

Ah, yes. We seem two have three copies of the #define in our tree, two under sys/, but it's not generally available. We could perhaps define it in cdefs.h, but independent of this change.

Done, please see https://reviews.freebsd.org/D56517

[rosenfeld_grumpf.hope-2000.org]

Review from emaste and markj.