On 10/1/2025 4:03 PM, Peter Xu wrote:
On Wed, Oct 01, 2025 at 03:07:23PM -0400, Steven Sistare wrote:
That does not make sense. It already does that, which is why I used it:
util/memfd.c
int qemu_memfd_create(const char *name, size_t size, bool hugetlb,
uint64_t hugetlbsize, unsigned int seals, Error **errp)
{
...
#ifdef CONFIG_LINUX
...
return mfd;
#else
error_setg_errno(errp, ENOSYS, "failed to create memfd");
#endif
return -1;
}
Did the windows build fail due to a different error?
https://gitlab.com/peterx/qemu/-/jobs/11566477462
See util/meson.build:
if host_os != 'windows'
...
util_ss.add(files('memfd.c'))
...
I see. So you were suggesting, as the minimal fix, to add ifdef CONFIG_LINUX at
the call site, not inside qemu_memfd_create, eg:
QEMUFile *cpr_exec_output(Error **errp)
{
#ifndef CONFIG_LINUX
error_setg(errp, "qemu_memfd_create requires CONFIG_LINUX");
return NULL;
#endif
int mfd = qemu_memfd_create(CPR_EXEC_STATE_NAME, 0, false, 0, 0, errp);
if (mfd < 0) {
return NULL;
}
return qemu_file_new_fd_output(mfd, CPR_EXEC_STATE_NAME);
}
Do you want me to re-submit a patch, or can you add the above to your tree?
If the latter, feel free to tweak it.
- Steve
