Ah, this is the new code.
Please reorder it to before wiring it up in mkfs, or even better just
merge the two patches as that makes reviewing easier.
> +++ b/mkfs/populate.c
> @@ -0,0 +1,287 @@
>
Please add the SPDX license identifier and your or your employers
copyright (whoever has the right to it) here.
> +static void fail(char *msg, int i)
> +{
> + fprintf(stderr, "%s: %s [%d - %s]\n", progname, msg, i, strerror(i));
This needs the _() treatment for localization.
> +static void writetimestamps(struct xfs_inode *ip, struct stat statbuf)
> +{
> + struct timespec64 ts;
> +
> + /* Copy timestamps from source file to destination inode.
> + * In order to not be influenced by our own access timestamp,
The usual kernel and xfprogs style for multi-line comments is:
/*
* Copy timestamps from source file to destination inode.
* ..
*/`
> +static void create_file(xfs_mount_t *mp, struct xfs_inode *pip,
We're (way too slowly) phasing out the use of typedefs for structs,
so the xfs_mount_t above should be struct xfs_mount.
> + xfs_inode_t *ip;
> + xfs_trans_t *tp;
> + tp = getres(mp, 0);
> + ppargs = newpptr(mp);
Similar for the xfs_inode_t and xfs_trans_t above and a few more later
in the patch.
Also please keep an empty line between delarations and the actual code
for clarity.
> + // copy over timestamps
Please use /* */ style comments.
> + if ((dir = opendir(cur_path)) == NULL) {
> + fail(_("cannot open input dir"), 1);
> + }
No need to use braces for single line statements.
> + while ((entry = readdir(dir)) != NULL) {
> + char link_target[PATH_MAX];
> + char path[PATH_MAX];
> + int error;
> + int fd = -1;
> + int flags;
> + int majdev;
> + int mindev;
> + int mode;
> + off_t len;
Can you factor this quite huge loop body into a helper function?
> index 0000000..e1b8587
> --- /dev/null
> +++ b/mkfs/populate.h
> @@ -0,0 +1,4 @@
> +#ifndef MKFS_POPULATE_H_
> +#define MKFS_POPULATE_H_
This also needs SPDX tag.
As you might have noticed this is really just nitpicking. The actual
logic looks good to me.
