Re:[ECOS] Re: cannot mount effs2 filesystem at boot

Re:[ECOS] Re: cannot mount effs2 filesystem at boot

[brix82]

---------- Initial Header -----------

From      : ecos-discuss-owner@ecos.sourceware.org
To          : "brix82@libero.it" brix82@libero.it
Cc          : "Andrew Lunn" andrew@lunn.ch,"ecos-discuss"
ecos-discuss@sources.redhat.com
Date      : Mon, 30 Aug 2004 16:36:40 +0200
Subject : [ECOS] Re: cannot mount effs2 filesystem at boot

> On Mon, Aug 30, 2004 at 04:10:23PM +0200, brix82@libero.it wrote:
> > I have configured ecos, now the device "/dev/flash1" points to the correct
> > location of the FIS directory:
> >
> > cdl_option CYGNUM_IO_FLASH_BLOCK_OFFSET_1 {
> >     # Flavor: data
> >     user_value 0x00300000
> > }
> >
> > And here is the FIS directory listing:
> >
> > RedBoot> fis list
> > Name              FLASH addr  Mem addr    Length      Entry point
> > RedBoot           0x50000000  0x50000000  0x00040000  0x00000000
> > ...
> > jffs              0x50300000  0x50300000  0x00080000  0x00050000
> >
> > Unfortunately, i can't write to the filesystem. I modified
> > /opt/ecos.massimo/ecos-2.0/packages/fs/jffs2/v2_0/src/os-ecos.h
> > and changed CONFIG_JFFS2_FS_DEBUG to 2. Here are the last few lines of the
output:
> >
> > ...
> > <7>erase_pending_list: 00000000 (used 00000000, dirty 00000000, wasted 00000000,
> > unchecked 00000000, free 00040000)
> > <7>erasable_pending_wbuf_list: empty
> > <7>free_list: empty
> > <7>bad_list: empty
> > <7>bad_used_list: empty
> > <7>Not rotating empty clean_list
> > <7>Not rotating empty very_dirty_list
> > <7>Not rotating empty dirty_list
> > <7>Not rotating empty erasable_list
> > <7>Rotating erase_pending_list by 0
> > <7>Erase block at front of erase_pending_list is at 00000000
> > <7>Not rotating empty free_list
> > <7>jffs2_read_super(): Getting root inode
> > <7>jffs2_read_inode(): inode->i_ino == 1
> > <7>Allocated inocache at 0x00050438
> > <7>jffs2_do_read_inode(): Creating inocache for root inode
> > <7>jffs2_do_read_inode_internal(): ino #1 nlink is 1
> > <7>jffs2_get_inode_nodes(): ino #1
> > <7>jffs2_read_inode() returning
> > <7>jffs2_read_super(): d_alloc_root()
> > <7>Starting erase of pending block 0x00000000
> > <7>Freeing all node refs for eraseblock offset 0x00000000
> > <7>Erase completed successfully at 0x00000000
> > <7>Verifying erase at 0x00000000
> > <7>Writing erased marker to block at 0x00000000
> > <7>jffs2_erase_pending_blocks completed
> > arenasize 66781112, freeblocks 1, totalallocated 1584, totalfree 66779508,
> > maxfree 66779508
> > <INFO>: reading directory /
> > <7>jffs2_readdir() for dir_i #1
> > <7>Dirent 0: ".", ino #1
> > <INFO>: entry              . [mode 00016f01 ino 000503b0 nlink 1 size 0]
> > <7>jffs2_readdir() for dir_i #1
> > <INFO>: entry             .. [mode 00016f01 ino 000503b0 nlink 1 size 0]
> > <7>jffs2_readdir() for dir_i #1
> > <INFO>: create file /foo size 202
> > jffs2_lookup()
> > jffs2_lookup()
> > <7>jffs2_create()
> > <7>jffs2_new_inode(): dir_i 1, mode 0x16f08
> > <7>Allocated inocache at 0x00050750
> > <7>jffs2_do_new_inode(): Assigned ino# 2
> > <7>jffs2_reserve_space(): Requested 0x44 bytes
> > <7>jffs2_reserve_space(): alloc sem got
> > <7>dirty size 0x00000000 + unchecked_size 0x00000000 < sector size 0x00040000,
> > returning -ENOSPC
>
> You have run out of memory.
>
> > I also have an application (a mini shell, this is my real interest) which should
> > browse the jffs2 filesystem image. In the main() function i do this:
> >
> > 	mount( "/dev/flash1", "/", "jffs2" );
> > 	while( 1 ) {
> > 		getcwd( pwd, 100 );
> > 		while( ( putchar(pwd[aptr++] ) ) != 0 );  // print PWD in the prompt
> >                 ...
> >        }
>
> Before you can getcwd() you first need to do a cddir() to /, otherwise
> the cwd is undefined and generally causes problems.
>
>         Andrew

Ok, i added some memory in the "Extra memory required by FLASH device drivers",
and now the shells works better, but not very well. For e
I built my filesystem image with mkfs.jffs2 1.39 for redhat with the following
command:

$ mkfs.jffs2 -o jffs.img -d jffs_tree/ -p 0x00400000 -e 0x00040000
$ ls -l jffs2.img
-rw-r--r--    1 root     root       262144  1 set 11:19 jffs.img

It produces a file of 256Kbytes (the erase block size), although i tried to pad
it to 4Mbytes. Is this ok? I think it's at the origin of the ENOSPC error i
still get when i do a fopen() to create a new file.
Here i try to create a file named "foo2" of size 1000 bytes (first 4 lines are
printed by the shell) :

> create foo2 1000
Creating file with name foo2
Creating file with size 1000
foo2
1000
jffs2_stat
icache_evict
init_dirsearch name = foo2
init_dirsearch dir = 152220
jffs2_find for path =foo2
find_entry
find_entry for name = foo2
jffs2_lookup()
find_entry got dir = 0
jffs2_open
icache_evict
init_dirsearch name = foo2
init_dirsearch dir = 152220
jffs2_find for path =foo2
find_entry
find_entry for name = foo2
jffs2_lookup()
find_entry got dir = 0
<7>jffs2_create()
<7>jffs2_new_inode(): dir_i 1, mode 0x16f08
malloc new_inode 152490 ####################################
<7>Allocated inocache at 0x00152518
<7>jffs2_do_new_inode(): Assigned ino# 6
<7>jffs2_add_ino_cache: Add 0x00152518 (ino #6)
<7>jffs2_reserve_space(): Requested 0x44 bytes
<7>jffs2_reserve_space(): alloc sem got
<7>dirty size 0x00000000 + unchecked_size 0x00000000 < sector size 0x00040000,
returning -ENOSPC
<7>jffs2_do_create(): reserved 0x1509c bytes
<7>jffs2_clear_inode(): ino #6 mode %o
make_bad_inode
free iput inode 152490 $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
<7>jffs2_clear_inode(): ino #6 mode %o
jffs2_getinfo
icache_evict
init_dirsearch name =
init_dirsearch dir = 152220
jffs2_find for path =
Error while opening file : No such entity>

It seems there is no space left on device, but i can't understand how to solve
the problem (or how to produce a jffs2 image with some free space, if this is
the solution).
Another strange but not important thing happens when i do the following sequence:

/> cd dir1
/dir1> cd ..
>

As you can see, when i return to the root directory the getcwd() function
doesn't give me the "/" string, but an empty one. This is not important however.
The real problem is to write to the filesystem.

Thanks in advance,

   Marco


--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Re: cannot mount effs2 filesystem at boot

[Andrew Lunn]

> > > And here is the FIS directory listing:
> > >
> > > RedBoot> fis list
> > > Name              FLASH addr  Mem addr    Length      Entry point
> > > RedBoot           0x50000000  0x50000000  0x00040000  0x00000000
> > > ...
> > > jffs              0x50300000  0x50300000  0x00080000  0x00050000
> > >

> I built my filesystem image with mkfs.jffs2 1.39 for redhat with the following
> command:
> 
> $ mkfs.jffs2 -o jffs.img -d jffs_tree/ -p 0x00400000 -e 0x00040000
> $ ls -l jffs2.img
> -rw-r--r--    1 root     root       262144  1 set 11:19 jffs.img
> 
> It produces a file of 256Kbytes (the erase block size), although i tried to pad
> it to 4Mbytes. Is this ok? 

Have you changed the fis entry and the other CDL configurations to
4Mbytes?

Also mkfs.jffs2 should produce a filesystem of 4Mbytes. It looks like
there is a bug in the parameter handling. Try:

mkfs.jffs2 -d jffs_tree --pad=4194304 -e 262144 -o jffs.img 

-p does not seem to be working.

        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Re: cannot mount effs2 filesystem at boot

[brix82]

---------- Initial Header -----------

From      : ecos-discuss-owner@ecos.sourceware.org
To          : "brix82@libero.it" brix82@libero.it
Cc          : "andrew" andrew@lunn.ch,"ecos-discuss" ecos-discuss@sources.redhat.com
Date      : Wed, 1 Sep 2004 12:31:20 +0200
Subject : Re: [ECOS] Re: cannot mount effs2 filesystem at boot

> > > > And here is the FIS directory listing:
> > > >
> > > > RedBoot> fis list
> > > > Name              FLASH addr  Mem addr    Length      Entry point
> > > > RedBoot           0x50000000  0x50000000  0x00040000  0x00000000
> > > > ...
> > > > jffs              0x50300000  0x50300000  0x00080000  0x00050000
> > > >
>
> > I built my filesystem image with mkfs.jffs2 1.39 for redhat with the following
> > command:
> >
> > $ mkfs.jffs2 -o jffs.img -d jffs_tree/ -p 0x00400000 -e 0x00040000
> > $ ls -l jffs2.img
> > -rw-r--r--    1 root     root       262144  1 set 11:19 jffs.img
> > 
> > It produces a file of 256Kbytes (the erase block size), although i tried to pad
> > it to 4Mbytes. Is this ok?
>
> Have you changed the fis entry and the other CDL configurations to
> 4Mbytes?
>
> Also mkfs.jffs2 should produce a filesystem of 4Mbytes. It looks like
> there is a bug in the parameter handling. Try:
>
> mkfs.jffs2 -d jffs_tree --pad=4194304 -e 262144 -o jffs.img
>
> -p does not seem to be working.
>
>         Andrew
>
> --
> Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
> and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss
>

In fact mkfs.jffs2 is bugged, and with the long format option all goes right.
But now there's another problem: i can't create big files because of write
errors. Do you think it's due to buggy hardware?
This is the result of the 'create' command:

/system> create wre.jar 1478774 0x100000                           
                              |
<5>Write of %zd bytes at 0x00000057 failed. returned 3014580, retlen %z d
<5>Write of %zd bytes at 0x000000f0 failed. returned 3080104, retlen %z d
<5>Write of %zd bytes at 0x000000de failed. returned 3211136, retlen %z d
<5>Write of %zd bytes at 0x000000cf failed. returned 3276784, retlen %z d
<5>Write of %zd bytes at 0x00000057 failed. returned 3342336, retlen %z d
<5>Write of %zd bytes at 0x00000130 failed. returned 2686972, retlen %z d
<5>Write of %zd bytes at 0x000000cd failed. returned 2752332, retlen %z d
<5>Write of %zd bytes at 0x000000f7 failed. returned 2817824, retlen %z d
<5>Write of %zd bytes at 0x00000126 failed. returned 2424688, retlen %z d
<5>Write of %zd bytes at 0x00000130 failed. returned 2490308, retlen %z d
<5>Write of %zd bytes at 0x000000f2 failed. returned 2555804, retlen %z d

And then the shell freezes. It seems to freeze on the close() call. When i reset
the board, i have a bunch of CRC errors.

<5>jffs2_get_inode_nodes(): CRC failed on node at 0x0031fff0: Read 0xffffffff,
calculated 0x18c9dceb
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x0030ff80: Read
0x96cc0ab3, calculated 0x3e2b0d6d
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x002effa8: Read
0x6b180c1d, calculated 0xb3109ff2
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x002dffb4: Read
0xf58f44e5, calculated 0xefd5c185
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x002aff20: Read
0xc7ab4f34, calculated 0x13c766c1
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x0029ff4c: Read
0xacba860d, calculated 0x67db399e
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x0026ff9c: Read
0xb9e796c4, calculated 0x420547ff
<5>jffs2_get_inode_nodes(): CRC failed on node at 0x0025ffc4: Read 0x0000000c,
calculated 0x902f562d
<5>jffs2_get_inode_nodes(): Data CRC failed on node at 0x0024ff70: Read
0x781ca088, calculated 0xb9023153
wre.jar          -rwxrwxrwx 1357736     0       1       1478774

I think this flash memory is really buggy... :(


--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: cannot mount effs2 filesystem at boot

[Andrew Lunn]

On Mon, Aug 30, 2004 at 04:10:23PM +0200, brix82@libero.it wrote:
> I have configured ecos, now the device "/dev/flash1" points to the correct
> location of the FIS directory:
> 
> cdl_option CYGNUM_IO_FLASH_BLOCK_OFFSET_1 {
>     # Flavor: data
>     user_value 0x00300000
> }
> 
> And here is the FIS directory listing:
> 
> RedBoot> fis list
> Name              FLASH addr  Mem addr    Length      Entry point
> RedBoot           0x50000000  0x50000000  0x00040000  0x00000000
> ...
> jffs              0x50300000  0x50300000  0x00080000  0x00050000
> 
> Unfortunately, i can't write to the filesystem. I modified
> /opt/ecos.massimo/ecos-2.0/packages/fs/jffs2/v2_0/src/os-ecos.h
> and changed CONFIG_JFFS2_FS_DEBUG to 2. Here are the last few lines of the output:
> 
> ...
> <7>erase_pending_list: 00000000 (used 00000000, dirty 00000000, wasted 00000000,
> unchecked 00000000, free 00040000)
> <7>erasable_pending_wbuf_list: empty
> <7>free_list: empty
> <7>bad_list: empty
> <7>bad_used_list: empty
> <7>Not rotating empty clean_list
> <7>Not rotating empty very_dirty_list
> <7>Not rotating empty dirty_list
> <7>Not rotating empty erasable_list
> <7>Rotating erase_pending_list by 0
> <7>Erase block at front of erase_pending_list is at 00000000
> <7>Not rotating empty free_list
> <7>jffs2_read_super(): Getting root inode
> <7>jffs2_read_inode(): inode->i_ino == 1
> <7>Allocated inocache at 0x00050438
> <7>jffs2_do_read_inode(): Creating inocache for root inode
> <7>jffs2_do_read_inode_internal(): ino #1 nlink is 1
> <7>jffs2_get_inode_nodes(): ino #1
> <7>jffs2_read_inode() returning
> <7>jffs2_read_super(): d_alloc_root()
> <7>Starting erase of pending block 0x00000000
> <7>Freeing all node refs for eraseblock offset 0x00000000
> <7>Erase completed successfully at 0x00000000
> <7>Verifying erase at 0x00000000
> <7>Writing erased marker to block at 0x00000000
> <7>jffs2_erase_pending_blocks completed
> arenasize 66781112, freeblocks 1, totalallocated 1584, totalfree 66779508,
> maxfree 66779508
> <INFO>: reading directory /
> <7>jffs2_readdir() for dir_i #1
> <7>Dirent 0: ".", ino #1
> <INFO>: entry              . [mode 00016f01 ino 000503b0 nlink 1 size 0]
> <7>jffs2_readdir() for dir_i #1
> <INFO>: entry             .. [mode 00016f01 ino 000503b0 nlink 1 size 0]
> <7>jffs2_readdir() for dir_i #1
> <INFO>: create file /foo size 202
> jffs2_lookup()
> jffs2_lookup()
> <7>jffs2_create()
> <7>jffs2_new_inode(): dir_i 1, mode 0x16f08
> <7>Allocated inocache at 0x00050750
> <7>jffs2_do_new_inode(): Assigned ino# 2
> <7>jffs2_reserve_space(): Requested 0x44 bytes
> <7>jffs2_reserve_space(): alloc sem got
> <7>dirty size 0x00000000 + unchecked_size 0x00000000 < sector size 0x00040000,
> returning -ENOSPC

You have run out of memory. 

> I also have an application (a mini shell, this is my real interest) which should
> browse the jffs2 filesystem image. In the main() function i do this:
> 
> 	mount( "/dev/flash1", "/", "jffs2" );
> 	while( 1 ) {
> 		getcwd( pwd, 100 );
> 		while( ( putchar(pwd[aptr++] ) ) != 0 );  // print PWD in the prompt
>                 ...
>        }

Before you can getcwd() you first need to do a cddir() to /, otherwise 
the cwd is undefined and generally causes problems.

        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Re: cannot mount effs2 filesystem at boot

[brix82]

I have configured ecos, now the device "/dev/flash1" points to the correct
location of the FIS directory:

cdl_option CYGNUM_IO_FLASH_BLOCK_OFFSET_1 {
    # Flavor: data
    user_value 0x00300000
}

And here is the FIS directory listing:

RedBoot> fis list
Name              FLASH addr  Mem addr    Length      Entry point
RedBoot           0x50000000  0x50000000  0x00040000  0x00000000
...
jffs              0x50300000  0x50300000  0x00080000  0x00050000

Unfortunately, i can't write to the filesystem. I modified
/opt/ecos.massimo/ecos-2.0/packages/fs/jffs2/v2_0/src/os-ecos.h
and changed CONFIG_JFFS2_FS_DEBUG to 2. Here are the last few lines of the output:

...
<7>erase_pending_list: 00000000 (used 00000000, dirty 00000000, wasted 00000000,
unchecked 00000000, free 00040000)
<7>erasable_pending_wbuf_list: empty
<7>free_list: empty
<7>bad_list: empty
<7>bad_used_list: empty
<7>Not rotating empty clean_list
<7>Not rotating empty very_dirty_list
<7>Not rotating empty dirty_list
<7>Not rotating empty erasable_list
<7>Rotating erase_pending_list by 0
<7>Erase block at front of erase_pending_list is at 00000000
<7>Not rotating empty free_list
<7>jffs2_read_super(): Getting root inode
<7>jffs2_read_inode(): inode->i_ino == 1
<7>Allocated inocache at 0x00050438
<7>jffs2_do_read_inode(): Creating inocache for root inode
<7>jffs2_do_read_inode_internal(): ino #1 nlink is 1
<7>jffs2_get_inode_nodes(): ino #1
<7>jffs2_read_inode() returning
<7>jffs2_read_super(): d_alloc_root()
<7>Starting erase of pending block 0x00000000
<7>Freeing all node refs for eraseblock offset 0x00000000
<7>Erase completed successfully at 0x00000000
<7>Verifying erase at 0x00000000
<7>Writing erased marker to block at 0x00000000
<7>jffs2_erase_pending_blocks completed
arenasize 66781112, freeblocks 1, totalallocated 1584, totalfree 66779508,
maxfree 66779508
<INFO>: reading directory /
<7>jffs2_readdir() for dir_i #1
<7>Dirent 0: ".", ino #1
<INFO>: entry              . [mode 00016f01 ino 000503b0 nlink 1 size 0]
<7>jffs2_readdir() for dir_i #1
<INFO>: entry             .. [mode 00016f01 ino 000503b0 nlink 1 size 0]
<7>jffs2_readdir() for dir_i #1
<INFO>: create file /foo size 202
jffs2_lookup()
jffs2_lookup()
<7>jffs2_create()
<7>jffs2_new_inode(): dir_i 1, mode 0x16f08
<7>Allocated inocache at 0x00050750
<7>jffs2_do_new_inode(): Assigned ino# 2
<7>jffs2_reserve_space(): Requested 0x44 bytes
<7>jffs2_reserve_space(): alloc sem got
<7>dirty size 0x00000000 + unchecked_size 0x00000000 < sector size 0x00040000,
returning -ENOSPC
<7>jffs2_do_create(): reserved 0x15e98 bytes
<7>jffs2_clear_inode(): ino #2 mode %o
<7>jffs2_clear_inode(): ino #2 mode %o
<FAIL>: open() returned -1 Unknown error
<FAIL>: write() returned -1 Bad file handle
<FAIL>: write() returned -1 Bad file handle
<FAIL>: write() returned -1 Bad file handle
...

What is wrong? I cannot find any reason for these errors.
(There are no 'lock' and 'unlock' commands in RedBoot, i guess it's not the
latest release. So it's not a locking problem)

I also have an application (a mini shell, this is my real interest) which should
browse the jffs2 filesystem image. In the main() function i do this:

	mount( "/dev/flash1", "/", "jffs2" );
	while( 1 ) {
		getcwd( pwd, 100 );
		while( ( putchar(pwd[aptr++] ) ) != 0 );  // print PWD in the prompt
                ...
       }

However, when i execute the binary i get errors:

RedBoot> go
<7>jffs2: read_super
<7>Allocating readbuf of 4096 bytes
<7>jffs2_scan_eraseblock(): Scanning block at 0x0
<5>jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000000: 0x2003
instead
<5>jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000004: 0x000c
instead
<5>jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00000008: 0xdc6d
instead
<7>Found empty flash at 0x00000010
<7>Empty flash ends normally at 0x00000400
<7>Fewer than %zd bytes (node header) left to end of buf. Reading 0xc at 0x00001000
<7>Found empty flash at 0x00000404
<7>Empty flash ends normally at 0x00001400
<7>Fewer than %zd bytes (node header) left to end of buf. Reading 0xc at 0x00001000
...

It seems to be looking for a filesystem at the address 0x0 instead of
0x50300000; but why did the fileio1 binary find it?

Thanks,

     Marco


--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss