[9fans] lib9p help

[9fans] lib9p help

[Gabriel Diaz]

Hello

i am using file/tree structures trying to do a read/write filesystem,
but i cannot make it use my walk functions and cannot create
directories or files with touch.

i read /sys/src/lib9p/srv.c and seems that in swalk():

if(r->fid->file)
    filewalk(r);
else if ( srv->walk1)
    walkandclone(r,oldwalk1,oldclone,srv);
else if ( srv->walk )
    srv->walk(r);
else
    sysfatal(...);

and firewalk(r) is always called.

what is the piece I'm missing? where is filled that r->fid->file ?

i have srv defined as:

srv fs = {
   .destroyfid = fsdestroyfid,
   .create=fscreate,
   .walk1=fswalk1,
   .clone=fsclone,
   .write=fswrite,
   .read=fsread,
   .remove=fsremove,
};



thanks, and sorry if this is obvious.

gabi

Re: [9fans] lib9p help

[Russ Cox]

> i am using file/tree structures trying to do a read/write filesystem,
> but i cannot make it use my walk functions and cannot create
> directories or files with touch.

If you are using file/tree structures, 
then you don't get to provide the walk function.
(Why would you need to?  Lib9p should know
the entire file tree and be able to do it for you.)

Russ

Re: [9fans] lib9p help

[Russ Cox]

> always fails when using lib9p. i though i need to provide walk1 and
> clone to create
> directories that doesn't exists. Using the chatty option all messages are walks
> when doing mkdirs abd when touching a file.

i don't believe that.  there should be Tcreate messages too.

/sys/src/lib9p/ramfs.c uses file/tree structures
(and allows creation of new files).  9p(2) says

          Create
               The create function must fill in both r->fid->qid and
               r->ofcall.qid on success.  When using file trees,
               create should allocate a new File with createfile; note
               that createfile may return nil (because, say, the file
               already exists).  If the create function is nil, srv
               behaves as though it were a function that always
               responded with the error ``create prohibited''.

russ

Re: [9fans] lib9p help

[Russ Cox]

> > /sys/src/lib9p/ramfs.c uses file/tree structures
> ramfs has it's own loop, may be it uses something like file/tree structures
> but the ramfs.c i have here doesn't have a  #include <9p.h>.

you are looking at /sys/src/cmd/ramfs.c and not /sys/src/lib9p/ramfs.c

> i have other tests with more fscalls implemented, but i should see
> here the Rwalk message and the Tcreate message?

yes, you should.  make sure you run your test with -c so
that the kernel will actually allow create in the root.

russ

Re: [9fans] lib9p help

[Gabriel Diaz]

Hello

that means that i cannot do an mkdir in a mounted fileserver while
using file/tree structures? that i need to generate the complete tree
inside the fileserver first?

something like
% fs -m /mnt/test
% mkdir /mnt/test/dir1

always fails when using lib9p. i though i need to provide walk1 and
clone to create
directories that doesn't exists. Using the chatty option all messages are walks
when doing mkdirs abd when touching a file.

it is better not to use lib9p for this? ( or file/free structures? )

thanks,

gabi


On 2/17/06, Russ Cox <rsc@swtch.com> wrote:
> > i am using file/tree structures trying to do a read/write filesystem,
> > but i cannot make it use my walk functions and cannot create
> > directories or files with touch.
>
> If you are using file/tree structures,
> then you don't get to provide the walk function.
> (Why would you need to?  Lib9p should know
> the entire file tree and be able to do it for you.)
>
> Russ
>
>

Re: [9fans] lib9p help

[Gabriel Diaz]

Hello,

i can see the Tcreate messages in ramfs, but not in my buggy test.

ramfs has it's own loop, may be it uses something like file/tree structures
but the ramfs.c i have here doesn't have a  #include <9p.h>.

using chatty i see that ramfs has a Rwalk messagfe with a newquid and
then the Tcreate
message, while in the test with 9p.h i never got it, so it fails.

i have other tests with more fscalls implemented, but i should see
here the Rwalk message and the Tcreate message?

thanks

gabi


my test fs.c is like:

/*
 * test1
 */

#include <u.h>
#include <libc.h>
#include <bio.h>
#include <auth.h>
#include <fcall.h>
#include <thread.h>
#include <9p.h>


Tree * tree;
File *ctlroot;

Srv fs;

static void
usage(void)
{
	fprint(2, "usage: fs [-abcC] [-m mtpt]\n");
	exits("usage");
}

void
main(int argc, char **argv)
{

	ulong flag;
	char *mtpt;

	flag = 0;
	mtpt = "/mnt/test";
	ARGBEGIN{
	case 'D':
		chatty9p++;
		break;
	case 'a':
		flag |= MAFTER;
		break;
	case 'b':
		flag |= MBEFORE;
		break;
	case 'c':
		flag |= MCREATE;
		break;
	case 'C':
		flag |= MCACHE;
		break;
	case 'm':
		mtpt = ARGF();
		break;
	default:
		usage();
		break;
	}ARGEND;

	if(argc != 0)
		usage();
	if ( flag == 0 )
		flag = MAFTER;

	tree = fs.tree = alloctree(getuser(), "sys", DMDIR|0775,nil);
	ctlroot = createfile(fs.tree->root,"ctl",getuser(),0666,nil);

	postmountsrv(&fs, nil, mtpt, flag);
	exits(0);
}


On 2/17/06, Russ Cox <rsc@swtch.com> wrote:
> > always fails when using lib9p. i though i need to provide walk1 and
> > clone to create
> > directories that doesn't exists. Using the chatty option all messages are walks
> > when doing mkdirs abd when touching a file.
>
> i don't believe that.  there should be Tcreate messages too.
>
> /sys/src/lib9p/ramfs.c uses file/tree structures
> (and allows creation of new files).  9p(2) says
>
>           Create
>                The create function must fill in both r->fid->qid and
>                r->ofcall.qid on success.  When using file trees,
>                create should allocate a new File with createfile; note
>                that createfile may return nil (because, say, the file
>                already exists).  If the create function is nil, srv
>                behaves as though it were a function that always
>                responded with the error ``create prohibited''.
>
> russ
>
>

Re: [9fans] lib9p help

[Gabriel Diaz]

hello

oh! i will take a look at that ramfs.

thanks

gabi


On 2/17/06, Russ Cox <rsc@swtch.com> wrote:
> > > /sys/src/lib9p/ramfs.c uses file/tree structures
> > ramfs has it's own loop, may be it uses something like file/tree structures
> > but the ramfs.c i have here doesn't have a  #include <9p.h>.
>
> you are looking at /sys/src/cmd/ramfs.c and not /sys/src/lib9p/ramfs.c
>
> > i have other tests with more fscalls implemented, but i should see
> > here the Rwalk message and the Tcreate message?
>
> yes, you should.  make sure you run your test with -c so
> that the kernel will actually allow create in the root.
>
> russ
>
>