Re: [PATCH] qnx6: convert qnx6 to use the new mount api

[Eric Sandeen <sandeen@redhat.com>]

On 2/29/24 1:13 PM, Bill O'Donnell wrote:
> Convert the qnx6 filesystem to use the new mount API.
> 
> Untested, since there is no qnx6 fs image readily available.
> 
> Signed-off-by: Bill O'Donnell <bodonnel@redhat.com>
> ---
>  fs/qnx6/inode.c | 119 +++++++++++++++++++++++++++++-------------------
>  1 file changed, 72 insertions(+), 47 deletions(-)
> 
> diff --git a/fs/qnx6/inode.c b/fs/qnx6/inode.c
> index a286c545717f..0df5a92a8b65 100644
> --- a/fs/qnx6/inode.c
> +++ b/fs/qnx6/inode.c
> @@ -19,11 +19,12 @@
>  #include <linux/buffer_head.h>
>  #include <linux/writeback.h>
>  #include <linux/statfs.h>
> -#include <linux/parser.h>
>  #include <linux/seq_file.h>
>  #include <linux/mount.h>

I *think* you can lose this include too but not a big deal.

>  #include <linux/crc32.h>
>  #include <linux/mpage.h>
> +#include <linux/fs_parser.h>
> +#include <linux/fs_context.h>
>  #include "qnx6.h"
>  
>  static const struct super_operations qnx6_sops;
> @@ -31,7 +32,7 @@ static const struct super_operations qnx6_sops;
>  static void qnx6_put_super(struct super_block *sb);
>  static struct inode *qnx6_alloc_inode(struct super_block *sb);
>  static void qnx6_free_inode(struct inode *inode);
> -static int qnx6_remount(struct super_block *sb, int *flags, char *data);
> +static int qnx6_reconfigure(struct fs_context *fc);
>  static int qnx6_statfs(struct dentry *dentry, struct kstatfs *buf);
>  static int qnx6_show_options(struct seq_file *seq, struct dentry *root);
>  
> @@ -40,7 +41,6 @@ static const struct super_operations qnx6_sops = {
>  	.free_inode	= qnx6_free_inode,
>  	.put_super	= qnx6_put_super,
>  	.statfs		= qnx6_statfs,
> -	.remount_fs	= qnx6_remount,
>  	.show_options	= qnx6_show_options,
>  };
>  
> @@ -54,10 +54,12 @@ static int qnx6_show_options(struct seq_file *seq, struct dentry *root)
>  	return 0;
>  }
>  
> -static int qnx6_remount(struct super_block *sb, int *flags, char *data)
> +static int qnx6_reconfigure(struct fs_context *fc)
>  {
> +	struct super_block *sb = fc->root->d_sb;
> +
>  	sync_filesystem(sb);
> -	*flags |= SB_RDONLY;
> +	fc->sb_flags |= SB_RDONLY;
>  	return 0;
>  }
>  
> @@ -222,35 +224,36 @@ enum {
>  	Opt_err
>  };
>  
> -static const match_table_t tokens = {
> -	{Opt_mmifs, "mmi_fs"},
> -	{Opt_err, NULL}
> +struct qnx6_context {
> +	unsigned long s_mount_opts;
> +};

s_mount_opts seems to be a write-only variable. It's set in 
qnx6_parse_param() but nothing ever reads it?

> +
> +static const struct fs_parameter_spec qnx6_param_spec[] = {
> +	fsparam_flag	("mmi_fs",	Opt_mmifs),
> +	{}
>  };
>  
> -static int qnx6_parse_options(char *options, struct super_block *sb)
> +static int qnx6_parse_param(struct fs_context *fc, struct fs_parameter *param)
>  {
> -	char *p;
> -	struct qnx6_sb_info *sbi = QNX6_SB(sb);
> -	substring_t args[MAX_OPT_ARGS];
> -
> -	if (!options)
> -		return 1;
> -
> -	while ((p = strsep(&options, ",")) != NULL) {
> -		int token;
> -		if (!*p)
> -			continue;
> -
> -		token = match_token(p, tokens, args);
> -		switch (token) {
> -		case Opt_mmifs:
> -			set_opt(sbi->s_mount_opt, MMI_FS);
> -			break;
> -		default:
> -			return 0;
> -		}
> +	struct qnx6_context *ctx = fc->fs_private;
> +	struct fs_parse_result result;
> +	int opt;
> +
> +	opt = fs_parse(fc, qnx6_param_spec, param, &result);
> +	if (opt < 0)
> +		return opt;
> +
> +	switch (opt) {
> +	case Opt_err:
> +		ctx->s_mount_opts |= result.uint_32;
> +		break;

Not sure what's going on here. Opt_err is not associated with any
valid mount option.

> +	case Opt_mmifs:
> +		ctx->s_mount_opts |= QNX6_MOUNT_MMI_FS;
> +		break;

This sets QNX6_MOUNT_MMI_FS into ctx->s_mount_opts but it looks like
nothing ever reads it back out of the context.

In qnx6_fill_super uptream, this handles the case where the mmi_fs option
was set:

         if (test_opt(s, MMI_FS)) {
                sb1 = qnx6_mmi_fill_super(s, silent);
                if (sb1)
                        goto mmi_success;
                else
                        goto outnobh;
        }       

Under the new mount api, the mmi_fs state needs to be saved in the
filesystem context when it's parsed, then tested here, to do the right
thing in fill_super when that option has been set.

> +	default:
> +		return -EINVAL;
>  	}
> -	return 1;
> +	return 0;
>  }
>  
>  static struct buffer_head *qnx6_check_first_superblock(struct super_block *s,
> @@ -293,22 +296,24 @@ static struct buffer_head *qnx6_check_first_superblock(struct super_block *s,
>  static struct inode *qnx6_private_inode(struct super_block *s,
>  					struct qnx6_root_node *p);
>  
> -static int qnx6_fill_super(struct super_block *s, void *data, int silent)
> +static int qnx6_fill_super(struct super_block *s, struct fs_context *fc)
>  {
>  	struct buffer_head *bh1 = NULL, *bh2 = NULL;
>  	struct qnx6_super_block *sb1 = NULL, *sb2 = NULL;
>  	struct qnx6_sb_info *sbi;
> +
>  	struct inode *root;
>  	const char *errmsg;
>  	struct qnx6_sb_info *qs;
>  	int ret = -EINVAL;
>  	u64 offset;
>  	int bootblock_offset = QNX6_BOOTBLOCK_SIZE;
> +	int silent = fc->sb_flags & SB_SILENT;
>  
> -	qs = kzalloc(sizeof(struct qnx6_sb_info), GFP_KERNEL);
> -	if (!qs)
> +	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
> +	if (!sbi)
>  		return -ENOMEM;
> -	s->s_fs_info = qs;
> +	s->s_fs_info = sbi;

Not sure what this change is for. The existing code is a little odd, it
allocates *qs as a struct qnx6_sb_info and assigns it to s->s_fs_info, then
reads it back out of s->s_fs_info via QNX6_SB(s) and assigns that to *sbi;
I don't see any real reason for that dance, it's 2 pointers w/ the same value.
 
But, with your change *qs is never initialized, and yet this:

outnobh:
        kfree(qs);

remains, which would free the uninitialized pointer variable.

I don't think any of the above needs to change for this conversion.

-Eric

>  	/* Superblock always is 512 Byte long */
>  	if (!sb_set_blocksize(s, QNX6_SUPERBLOCK_SIZE)) {
> @@ -316,11 +321,6 @@ static int qnx6_fill_super(struct super_block *s, void *data, int silent)
>  		goto outnobh;
>  	}
>  
> -	/* parse the mount-options */
> -	if (!qnx6_parse_options((char *) data, s)) {
> -		pr_err("invalid mount options.\n");
> -		goto outnobh;
> -	}
>  	if (test_opt(s, MMI_FS)) {
>  		sb1 = qnx6_mmi_fill_super(s, silent);
>  		if (sb1)
> @@ -632,18 +632,43 @@ static void destroy_inodecache(void)
>  	kmem_cache_destroy(qnx6_inode_cachep);
>  }
>  
> -static struct dentry *qnx6_mount(struct file_system_type *fs_type,
> -	int flags, const char *dev_name, void *data)
> +static int qnx6_get_tree(struct fs_context *fc)
>  {
> -	return mount_bdev(fs_type, flags, dev_name, data, qnx6_fill_super);
> +	return get_tree_bdev(fc, qnx6_fill_super);
> +}
> +
> +static void qnx6_free_fc(struct fs_context *fc)
> +{
> +	kfree(fc->fs_private);
> +}
> +
> +static const struct fs_context_operations qnx6_context_ops = {
> +	.parse_param	= qnx6_parse_param,
> +	.get_tree	= qnx6_get_tree,
> +	.reconfigure	= qnx6_reconfigure,
> +	.free		= qnx6_free_fc,
> +};
> +
> +static int qnx6_init_fs_context(struct fs_context *fc)
> +{
> +	struct qnx6_context *ctx;
> +
> +	ctx = kzalloc(sizeof(struct qnx6_context), GFP_KERNEL);
> +	if (!ctx)
> +		return -ENOMEM;
> +	fc->ops = &qnx6_context_ops;
> +	fc->fs_private = ctx;
> +
> +	return 0;
>  }
>  
>  static struct file_system_type qnx6_fs_type = {
> -	.owner		= THIS_MODULE,
> -	.name		= "qnx6",
> -	.mount		= qnx6_mount,
> -	.kill_sb	= kill_block_super,
> -	.fs_flags	= FS_REQUIRES_DEV,
> +	.owner			= THIS_MODULE,
> +	.name			= "qnx6",
> +	.kill_sb		= kill_block_super,
> +	.fs_flags		= FS_REQUIRES_DEV,
> +	.init_fs_context	= qnx6_init_fs_context,
> +	.parameters		= qnx6_param_spec,
>  };
>  MODULE_ALIAS_FS("qnx6");
>