Re: [PATCH 1/3] 9p: add support for root file systems

Re: [PATCH 1/3] 9p: add support for root file systems

[Dominique Martinet]

Why has this only been sent to direct maintainers and linux-doc@vger ?
Please resend to at least v9fs-developer@sf, linux-fs@vger and whoever
get_maintainer.pl deems appropriate for init/do_mounts.c changes
(added the two lists to this mail)


As a whole for the series: I'm personally not sure I'd encourage this,
it can currently be done with an initrd if someone cares enough and if
we're going to add all remote filesystems that way there's going to be
no end to it.

That being said, I'm not 100% opposed to it if there is demand; but I'll
expect you to help with whatever bug reports people come up with using
this option.


Changbin Du wrote on Wed, May 05, 2021 at 08:07:46PM +0800:
> diff --git a/fs/9p/v9fsroot.c b/fs/9p/v9fsroot.c
> new file mode 100644
> index 000000000000..7dd91cc3814f
> --- /dev/null
> +++ b/fs/9p/v9fsroot.c
> @@ -0,0 +1,64 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * SMB root file system support

copy-paste error from fs/cifs/cifsroot.c

> + *
> + * Copyright (c) 2021 Changbin Du <changbin.du@gmail.com>
> + */
> +#include <linux/init.h>
> +#include <linux/fs.h>
> +#include <linux/types.h>
> +#include <linux/ctype.h>
> +#include <linux/string.h>
> +#include <linux/root_dev.h>
> +#include <linux/kernel.h>
> +
> +static char root_dev[2048] __initdata = "";
> +static char root_opts[1024] __initdata = "";
> +
> +/* v9fsroot=<path>[,options] */
> +static int __init v9fs_root_setup(char *line)
> +{
> +	char *s;
> +	int len;
> +
> +	ROOT_DEV = Root_V9FS;
> +
> +	if (strlen(line) >= 1) {
> +		/* make s point to ',' or '\0' at end of line */
> +		s = strchrnul(line, ',');
> +		/* len is strlen(unc) + '\0' */
> +		len = s - line + 1;
> +		if (len > sizeof(root_dev)) {
> +			printk(KERN_ERR "Root-V9FS: path too long\n");

9p has its own p9_debug helper; it's a bit awkward because even
P9_DEBUG_ERROR right now is not displayed by default so I'm not against
pr_err or pr_warn for important messages (I'd like to replace all such
messages by either at some point), but at least stick to pr_xyz(...)
rather than printk(KERN_XYZ...)

> +			return 1;
> +		}
> +		strlcpy(root_dev, line, len);
> +		if (*s) {
> +			int n = snprintf(root_opts,
> +					 sizeof(root_opts), "%s,%s",
> +					 DEFAULT_MNT_OPTS, s + 1);

Did you actually run with this? DEFAULT_MNT_OPTS is not defined here.

> +			if (n >= sizeof(root_opts)) {
> +				printk(KERN_ERR "Root-V9FS: mount options string too long\n");
> +				root_opts[sizeof(root_opts)-1] = '\0';
> +				return 1;
> +			}
> +		}
> +	}
> +
> +	return 1;

I'm also surprised this only ever returns 1, the cifs codes does the
same but I'd be surprised the mount can work without a tag so there
certainly should be some difference between working and not working?

I'd also expect ROOT_DEV to only be set on success, this doesn't make
sense to trigger 9p mounting mechanisms with something that can't be
mounted, leading to a long timeout (5+10+15+30*2 seconds) before a hard
error when the hard error can be given right away.

> +}
> +
> +__setup("v9fsroot=", v9fs_root_setup);
> +
> +int __init v9fs_root_data(char **dev, char **opts)
> +{
> +	if (!root_dev[0]) {
> +		printk(KERN_ERR "Root-V9FS: no rootdev specified\n");
> +		return -1;
> +	}
> +
> +	*dev = root_dev;
> +	*opts = root_opts;
> +
> +	return 0;
> +}
> diff --git a/include/linux/root_dev.h b/include/linux/root_dev.h
> index 4e78651371ba..becd0ee2ff87 100644
> --- a/include/linux/root_dev.h
> +++ b/include/linux/root_dev.h
> @@ -9,6 +9,7 @@
>  enum {
>  	Root_NFS = MKDEV(UNNAMED_MAJOR, 255),
>  	Root_CIFS = MKDEV(UNNAMED_MAJOR, 254),
> +	Root_V9FS = MKDEV(UNNAMED_MAJOR, 253),
>  	Root_RAM0 = MKDEV(RAMDISK_MAJOR, 0),
>  	Root_RAM1 = MKDEV(RAMDISK_MAJOR, 1),
>  	Root_FD0 = MKDEV(FLOPPY_MAJOR, 0),
> diff --git a/init/do_mounts.c b/init/do_mounts.c
> index a78e44ee6adb..60af89983a6b 100644
> --- a/init/do_mounts.c
> +++ b/init/do_mounts.c
> @@ -287,6 +287,8 @@ dev_t name_to_dev_t(const char *name)
>  		return Root_NFS;
>  	if (strcmp(name, "/dev/cifs") == 0)
>  		return Root_CIFS;
> +	if (strcmp(name, "/dev/v9fs") == 0)
> +		return Root_V9FS;
>  	if (strcmp(name, "/dev/ram") == 0)
>  		return Root_RAM0;
>  #ifdef CONFIG_BLOCK
> @@ -536,6 +538,43 @@ static int __init mount_cifs_root(void)
>  }
>  #endif
>  
> +#ifdef CONFIG_9P_FS_ROOT
> +
> +extern int v9fs_root_data(char **dev, char **opts);
> +
> +#define V9FSROOT_TIMEOUT_MIN	5
> +#define V9FSROOT_TIMEOUT_MAX	30
> +#define V9FSROOT_RETRY_MAX	5
> +
> +static int __init mount_v9fs_root(void)
> +{
> +	char *root_dev, *root_data;
> +	unsigned int timeout;
> +	int try, err;
> +
> +	err = v9fs_root_data(&root_dev, &root_data);
> +	if (err != 0)
> +		return 0;
> +
> +	timeout = V9FSROOT_TIMEOUT_MIN;
> +	for (try = 1; ; try++) {
> +		err = do_mount_root(root_dev, "9p",
> +				    root_mountflags, root_data);
> +		if (err == 0)
> +			return 1;
> +		if (try > V9FSROOT_RETRY_MAX)
> +			break;
> +
> +		/* Wait, in case the server refused us immediately */
> +		ssleep(timeout);
> +		timeout <<= 1;
> +		if (timeout > V9FSROOT_TIMEOUT_MAX)
> +			timeout = V9FSROOT_TIMEOUT_MAX;
> +	}
> +	return 0;
> +}
> +#endif
> +
>  void __init mount_root(void)
>  {
>  #ifdef CONFIG_ROOT_NFS
> @@ -552,6 +591,13 @@ void __init mount_root(void)
>  		return;
>  	}
>  #endif
> +#ifdef CONFIG_9P_FS_ROOT
> +	if (ROOT_DEV == Root_V9FS) {
> +		if (!mount_v9fs_root())
> +			printk(KERN_ERR "VFS: Unable to mount root fs via 9p.\n");
> +		return;
> +	}
> +#endif
>  #ifdef CONFIG_BLOCK
>  	{
>  		int err = create_dev("/dev/root", ROOT_DEV);
-- 
Dominique

Re: [PATCH 1/3] 9p: add support for root file systems

[Enrico Weigelt, metux IT consult]

On 05.05.21 14:50, Dominique Martinet wrote:

> As a whole for the series: I'm personally not sure I'd encourage this,
> it can currently be done with an initrd if someone cares enough and if
> we're going to add all remote filesystems that way there's going to be
> no end to it.

personally, I'd be *very* interested in that, especially for specially
tailored vm grid workloads. please keep me in the loop.


--mtx

-- 
---
Hinweis: unverschlüsselte E-Mails können leicht abgehört und manipuliert
werden ! Für eine vertrauliche Kommunikation senden Sie bitte ihren
GPG/PGP-Schlüssel zu.
---
Enrico Weigelt, metux IT consult
Free software and Linux embedded engineering
info@metux.net -- +49-151-27565287

Re: [PATCH 1/3] 9p: add support for root file systems

[Changbin Du]

On Wed, May 05, 2021 at 09:50:50PM +0900, Dominique Martinet wrote:
> Why has this only been sent to direct maintainers and linux-doc@vger ?
> Please resend to at least v9fs-developer@sf, linux-fs@vger and whoever
> get_maintainer.pl deems appropriate for init/do_mounts.c changes
> (added the two lists to this mail)
>
I made a typing mistake and then...

> 
> As a whole for the series: I'm personally not sure I'd encourage this,
> it can currently be done with an initrd if someone cares enough and if
> we're going to add all remote filesystems that way there's going to be
> no end to it.
> 
> That being said, I'm not 100% opposed to it if there is demand; but I'll
> expect you to help with whatever bug reports people come up with using
> this option.
> 
No problem, I can help with any issues related to this. It is worth doint this
especially for some virtulizaion scenarioes.

> 
> Changbin Du wrote on Wed, May 05, 2021 at 08:07:46PM +0800:
> > diff --git a/fs/9p/v9fsroot.c b/fs/9p/v9fsroot.c
> > new file mode 100644
> > index 000000000000..7dd91cc3814f
> > --- /dev/null
> > +++ b/fs/9p/v9fsroot.c
> > @@ -0,0 +1,64 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * SMB root file system support
> 
> copy-paste error from fs/cifs/cifsroot.c
> 
Sorry, I'll fix it.

> > + *
> > + * Copyright (c) 2021 Changbin Du <changbin.du@gmail.com>
> > + */
> > +#include <linux/init.h>
> > +#include <linux/fs.h>
> > +#include <linux/types.h>
> > +#include <linux/ctype.h>
> > +#include <linux/string.h>
> > +#include <linux/root_dev.h>
> > +#include <linux/kernel.h>
> > +
> > +static char root_dev[2048] __initdata = "";
> > +static char root_opts[1024] __initdata = "";
> > +
> > +/* v9fsroot=<path>[,options] */
> > +static int __init v9fs_root_setup(char *line)
> > +{
> > +	char *s;
> > +	int len;
> > +
> > +	ROOT_DEV = Root_V9FS;
> > +
> > +	if (strlen(line) >= 1) {
> > +		/* make s point to ',' or '\0' at end of line */
> > +		s = strchrnul(line, ',');
> > +		/* len is strlen(unc) + '\0' */
> > +		len = s - line + 1;
> > +		if (len > sizeof(root_dev)) {
> > +			printk(KERN_ERR "Root-V9FS: path too long\n");
> 
> 9p has its own p9_debug helper; it's a bit awkward because even
> P9_DEBUG_ERROR right now is not displayed by default so I'm not against
> pr_err or pr_warn for important messages (I'd like to replace all such
> messages by either at some point), but at least stick to pr_xyz(...)
> rather than printk(KERN_XYZ...)
> 
okay, will use pr_err() instead then.

> > +			return 1;
> > +		}
> > +		strlcpy(root_dev, line, len);
> > +		if (*s) {
> > +			int n = snprintf(root_opts,
> > +					 sizeof(root_opts), "%s,%s",
> > +					 DEFAULT_MNT_OPTS, s + 1);
> 
> Did you actually run with this? DEFAULT_MNT_OPTS is not defined here.
> 
oops, my bad that I forgot to test the latest change.

> > +			if (n >= sizeof(root_opts)) {
> > +				printk(KERN_ERR "Root-V9FS: mount options string too long\n");
> > +				root_opts[sizeof(root_opts)-1] = '\0';
> > +				return 1;
> > +			}
> > +		}
> > +	}
> > +
> > +	return 1;
> 
> I'm also surprised this only ever returns 1, the cifs codes does the
> same but I'd be surprised the mount can work without a tag so there
> certainly should be some difference between working and not working?
>
I think this returns 1 just means the "v9fsroot=" kernel parameter is handled
so do not need further processing. But this doesn't mean the mount options can
work.

> I'd also expect ROOT_DEV to only be set on success, this doesn't make
> sense to trigger 9p mounting mechanisms with something that can't be
> mounted, leading to a long timeout (5+10+15+30*2 seconds) before a hard
> error when the hard error can be given right away.
> 
Sure, will only set ROOT_DEV after basic option checking.

> > +}
> > +
> > +__setup("v9fsroot=", v9fs_root_setup);
> > +
> > +int __init v9fs_root_data(char **dev, char **opts)
> > +{
> > +	if (!root_dev[0]) {
> > +		printk(KERN_ERR "Root-V9FS: no rootdev specified\n");
> > +		return -1;
> > +	}
> > +
> > +	*dev = root_dev;
> > +	*opts = root_opts;
> > +
> > +	return 0;
> > +}
> > diff --git a/include/linux/root_dev.h b/include/linux/root_dev.h
> > index 4e78651371ba..becd0ee2ff87 100644
> > --- a/include/linux/root_dev.h
> > +++ b/include/linux/root_dev.h
> > @@ -9,6 +9,7 @@
> >  enum {
> >  	Root_NFS = MKDEV(UNNAMED_MAJOR, 255),
> >  	Root_CIFS = MKDEV(UNNAMED_MAJOR, 254),
> > +	Root_V9FS = MKDEV(UNNAMED_MAJOR, 253),
> >  	Root_RAM0 = MKDEV(RAMDISK_MAJOR, 0),
> >  	Root_RAM1 = MKDEV(RAMDISK_MAJOR, 1),
> >  	Root_FD0 = MKDEV(FLOPPY_MAJOR, 0),
> > diff --git a/init/do_mounts.c b/init/do_mounts.c
> > index a78e44ee6adb..60af89983a6b 100644
> > --- a/init/do_mounts.c
> > +++ b/init/do_mounts.c
> > @@ -287,6 +287,8 @@ dev_t name_to_dev_t(const char *name)
> >  		return Root_NFS;
> >  	if (strcmp(name, "/dev/cifs") == 0)
> >  		return Root_CIFS;
> > +	if (strcmp(name, "/dev/v9fs") == 0)
> > +		return Root_V9FS;
> >  	if (strcmp(name, "/dev/ram") == 0)
> >  		return Root_RAM0;
> >  #ifdef CONFIG_BLOCK
> > @@ -536,6 +538,43 @@ static int __init mount_cifs_root(void)
> >  }
> >  #endif
> >  
> > +#ifdef CONFIG_9P_FS_ROOT
> > +
> > +extern int v9fs_root_data(char **dev, char **opts);
> > +
> > +#define V9FSROOT_TIMEOUT_MIN	5
> > +#define V9FSROOT_TIMEOUT_MAX	30
> > +#define V9FSROOT_RETRY_MAX	5
> > +
> > +static int __init mount_v9fs_root(void)
> > +{
> > +	char *root_dev, *root_data;
> > +	unsigned int timeout;
> > +	int try, err;
> > +
> > +	err = v9fs_root_data(&root_dev, &root_data);
> > +	if (err != 0)
> > +		return 0;
> > +
> > +	timeout = V9FSROOT_TIMEOUT_MIN;
> > +	for (try = 1; ; try++) {
> > +		err = do_mount_root(root_dev, "9p",
> > +				    root_mountflags, root_data);
> > +		if (err == 0)
> > +			return 1;
> > +		if (try > V9FSROOT_RETRY_MAX)
> > +			break;
> > +
> > +		/* Wait, in case the server refused us immediately */
> > +		ssleep(timeout);
> > +		timeout <<= 1;
> > +		if (timeout > V9FSROOT_TIMEOUT_MAX)
> > +			timeout = V9FSROOT_TIMEOUT_MAX;
> > +	}
> > +	return 0;
> > +}
> > +#endif
> > +
> >  void __init mount_root(void)
> >  {
> >  #ifdef CONFIG_ROOT_NFS
> > @@ -552,6 +591,13 @@ void __init mount_root(void)
> >  		return;
> >  	}
> >  #endif
> > +#ifdef CONFIG_9P_FS_ROOT
> > +	if (ROOT_DEV == Root_V9FS) {
> > +		if (!mount_v9fs_root())
> > +			printk(KERN_ERR "VFS: Unable to mount root fs via 9p.\n");
> > +		return;
> > +	}
> > +#endif
> >  #ifdef CONFIG_BLOCK
> >  	{
> >  		int err = create_dev("/dev/root", ROOT_DEV);
> -- 
> Dominique

-- 
Cheers,
Changbin Du