Return -EUCLEAN on read when a bitflip was detected and corrected, so the
clients can react and eventually copy the affected block to a spare one.
Make all in kernel users aware of the change.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Hopefully the last iteration on this!
The handling of out of band data on NAND was accompanied by tons of fruitless
discussions and halfarsed patches to make it work for a particular
problem. Sufficiently annoyed by I all those "I know it better" mails and the
resonable amount of discarded "it solves my problem" patches, I finally decided
to go for the big rework. After removing the _ecc variants of mtd read/write
functions the solution to satisfy the various requirements was to refactor the
read/write _oob functions in mtd.
The major change is that read/write_oob now takes a pointer to an operation
descriptor structure "struct mtd_oob_ops".instead of having a function with at
least seven arguments.
read/write_oob which should probably renamed to a more descriptive name, can do
the following tasks:
- read/write out of band data
- read/write data content and out of band data
- read/write raw data content and out of band data (ecc disabled)
struct mtd_oob_ops has a mode field, which determines the oob handling mode.
Aside of the MTD_OOB_RAW mode, which is intended to be especially for
diagnostic purposes and some internal functions e.g. bad block table creation,
the other two modes are for mtd clients:
MTD_OOB_PLACE puts/gets the given oob data exactly to/from the place which is
described by the ooboffs and ooblen fields of the mtd_oob_ops strcuture. It's
up to the caller to make sure that the byte positions are not used by the ECC
placement algorithms.
MTD_OOB_AUTO puts/gets the given oob data automaticaly to/from the places in
the out of band area which are described by the oobfree tuples in the ecclayout
data structre which is associated to the devicee.
The decision whether data plus oob or oob only handling is done depends on the
setting of the datbuf member of the data structure. When datbuf == NULL then
the internal read/write_oob functions are selected, otherwise the read/write
data routines are invoked.
Tested on a few platforms with all variants. Please be aware of possible
regressions for your particular device / application scenario
Disclaimer: Any whining will be ignored from those who just contributed "hot
air blurb" and never sat down to tackle the underlying problem of the mess in
the NAND driver grown over time and the big chunk of work to fix up the
existing users. The problem was not the holiness of the existing MTD
interfaces. The problems was the lack of time to go for the big overhaul. It's
easy to add more mess to the existing one, but it takes alot of effort to go
for a real solution.
Improvements and bugfixes are welcome!
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Most of those macros are unused and the used ones just obfuscate
the code. Remove them and fixup all users.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The nand_oobinfo structure is not fitting the newer error correction
demands anymore. Replace it by struct nand_ecclayout and fixup the users
all over the place. Keep the nand_oobinfo based ioctl for user space
compability reasons.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The info structure for out of band data was copied into
the mtd structure. Make it a pointer and remove the ability
to set it from userspace. The position of ecc bytes is
defined by the hardware and should not be changed by software.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This allows us to drop another pointer from the struct jffs2_raw_node_ref,
shrinking it to 8 bytes on 32-bit machines (if the TEST_TOTLEN) paranoia
check is turned off, which will be committed soon).
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Print wasted_size in scanned eraseblocks, print range correctly for
summary dirent and inode entries.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Preallocation of refs is shortly going to be a per-eraseblock thing,
rather than per-filesystem. Add the required argument to the function.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
One more place where we were changing the accounting info without
actually allocating a ref for the lost space...
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Random unthinking 'cleanup' caused debug messages like this:
Obsoleting node at 0x0006daf4 of len 0x3a4: <7>Dirtying
If messages are continuation of an existing line, they don't need
to be prefixed with KERN_DEBUG.
THINK. Or you will be replaced by a small shell script.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
It looks like metapage_releasepage was making in invalid assumption that
the releasepage method would not be called on a dirty page. Instead of
issuing a warning and releasing the metapage, it should return 0, indicating
that the private data for the page cannot be released.
I also realized that metapage_releasepage had the return code all wrong. If
it is successful in releasing the private data, it should return 1, otherwise
it needs to return 0.
Lastly, there is no need to call wait_on_page_writeback, since
try_to_release_page will not call us with a page in writback state.
Signed-off-by: Dave Kleikamp <shaggy@austin.ibm.com>
As the first step towards eliminating the ref->next_phys member and saving
memory by using an _array_ of struct jffs2_raw_node_ref per eraseblock,
stop the write functions from allocating their own refs; have them just
_reserve_ the appropriate number instead. Then jffs2_link_node_ref() can
just fill them in.
Use a linked list of pre-allocated refs in the superblock, for now. Once
we switch to an array, it'll just be a case of extending that array.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Else a subsequent bio_clone might make a mess.
Signed-off-by: Neil Brown <neilb@suse.de>
Cc: "Don Dupuis" <dondster@gmail.com>
Acked-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Both cause the 'entries' count in the export cache to be non-zero at module
removal time, so unregistering that cache fails and results in an oops.
1/ exp_pseudoroot (used for NFSv4 only) leaks a reference to an export
entry.
2/ sunrpc_cache_update doesn't increment the entries count when it adds
an entry.
Thanks to "david m. richter" <richterd@citi.umich.edu> for triggering the
problem and finding one of the bugs.
Cc: "david m. richter" <richterd@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
MTD clients are agnostic of FLASH which needs ECC suppport.
Remove the functions and fixup the callers.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The writev based write buffer implementation was far to complex as
in most use cases the write buffer had to be handled anyway.
Simplify the write buffer handling and use mtd->write instead.
From extensive testing no performance impact has been noted.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
We don't need the upper layers to deal with the physical offset. It's
_always_ c->nextblock->offset + c->sector_size - c->nextblock->free_size
so we might as well just let the actual write functions deal with that.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
o Add a flag MTD_BIT_WRITEABLE for devices that allow single bits to be
cleared.
o Replace MTD_PROGRAM_REGIONS with a cleared MTD_BIT_WRITEABLE flag for
STMicro and Intel Sibley flashes with internal ECC. Those flashes
disallow clearing of single bits, unlike regular NOR flashes, so the
new flag models their behaviour better.
o Remove MTD_ECC. After the STMicro/Sibley merge, this flag is only set
and never checked.
Signed-off-by: Joern Engel <joern@wh.fh-wedel.de>
In 2002, STMicro started producing NOR flashes with internal ECC protection
for small blocks (8 or 16 bytes). Support for those flashes was added by me.
In 2005, Intel Sibley flashes copied this strategy and Nico added support for
those. Merge the code for both.
Signed-off-by: Joern Engel <joern@wh.fh-wedel.de>
At least two flashes exists that have the concept of a minimum write unit,
similar to NAND pages, but no other NAND characteristics. Therefore, rename
the minimum write unit to "writesize" for all flashes, including NAND.
Signed-off-by: Joern Engel <joern@wh.fh-wedel.de>
We'll be using a proper list of nodes in the jffs2_xattr_datum and
jffs2_xattr_ref structures, because the existing code to overwrite
them is just broken. Put it in the common part at the front of the
structure which is shared with the jffs2_inode_cache, so that the
jffs2_link_node_ref() function can do the right thing.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
In a couple of places, we assume that what's at the end of the
->next_in_ino list is a struct jffs2_inode_cache. Let's check
for that, since we expect it to change soon.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Let's avoid the potential for forgetting to set ref->next_in_ino, by doing
it within jffs2_link_node_ref() instead.
This highlights the ugliness of what we're currently doing with
xattr_datum and xattr_ref structures -- we should find a nicer way of
dealing with that.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
When filing REF_OBSOLETE nodes, we'd add their size to the global
'dirty_size' count, but then to the eraseblock's 'used_size' count.
That's not clever.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Don't reassign to watch. If idr_find() returns NULL, then
put_inotify_watch() will choke.
Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Cc: John McCutchan <john@johnmccutchan.com>
Cc: Robert Love <rlove@rlove.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
While doing some inotify stress testing, I hit the following race. In
inotify_release(), it's possible for a watch to be removed from the lists
in between dropping dev->mutex and taking inode->inotify_mutex. The
reference we hold prevents the watch from being freed, but not from being
removed.
Checking the dev's idr mapping will prevent a double list_del of the
same watch.
Signed-off-by: Amy Griffis <amy.griffis@hp.com>
Acked-by: John McCutchan <john@johnmccutchan.com>
Cc: Robert Love <rml@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Bernd Schmidt points out that binfmt_flat is now leaving the exec file open
while the application runs. This offsets all the application's fd numbers.
We should have closed the file within exec(), not at exit()-time.
But there doesn't seem to be a lot of point in doing all this just to avoid
going over RLIMIT_NOFILE by one fd for a few microseconds. So take the EMFILE
checking out again. This will cause binfmt_flat to again fail LTP's
exec-should-return-EMFILE-when-fdtable-is-full test. That test appears to be
wrong anyway - Open Group specs say nothing about exec() returning EMFILE.
Cc: Bernd Schmidt <bernd.schmidt@analog.com>
Cc: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Assigning the result of posix_acl_to_xattr() to an unsigned data type
(size/size_t) obscures possible errors.
Coverity CID: 1206.
Signed-off-by: Florin Malita <fmalita@gmail.com>
Acked-by: NeilBrown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Address a problem found when a Linux NFS server uses the "subtree_check"
export option.
The "subtree_check" NFS export option was designed to prohibit a client
from using a file handle for which it should not have permission. The
algorithm used is to ensure that the entire path to the file being
referenced is accessible to the user attempting to use the file handle. If
some part of the path is not accessible, then the operation is aborted and
the appropriate version of ESTALE is returned to the NFS client.
The error, ESTALE, is unfortunate in that it causes NFS clients to make
certain assumptions about the continued existence of the file. They assume
that the file no longer exists and refuse to attempt to access it again.
In this case, the file really does exist, but access was denied by the
server for a particular user.
A better error to return would be an EACCES sort of error. This would
inform the client that the particular operation that it was attempting was
not allowed, without the nasty side effects of the ESTALE error.
Signed-off-by: Peter Staubach <staubach@redhat.com>
Acked-By: NeilBrown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Functions compat_nfs_svc_trans, compat_nfs_clnt_trans,
compat_nfs_exp_trans, compat_nfs_getfd_trans and compat_nfs_getfs_trans,
which are called by compat_sys_nfsservctl(fs/compat.c), don't handle the
return value of access_ok properly. access_ok return 1 when the addr is
valid, and 0 when it's not, but these functions have the reversed
understanding. When the address is valid, they always return -EFAULT to
compat_sys_nfsservctl.
An example is to run /usr/sbin/rpc.nfsd(32bit program on Power5). It
doesn't function as expected. strace showes that nfsservctl returns
-EFAULT.
The patch fixes this by correcting the error handling on the return value
of access_ok in the five functions.
Signed-off-by: Lin Feng Shen <shenlinf@cn.ibm.com>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Acked-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Well, almost. We'll actually keep a 'TEST_TOTLEN' macro set for now, and keep
doing some paranoia checks to make sure it's all working correctly. But if
TEST_TOTLEN is unset, the size of struct jffs2_raw_node_ref drops from 16
bytes to 12 on 32-bit machines. That's a saving of about half a megabyte of
memory on the OLPC prototype board, with 125K or so nodes in its 512MiB of
flash.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
We can't use jffs2_scan_dirty_space() because it doesn't do any locking; it's
only for use at scan time -- hence the 'scan' in the name.
Also, don't allocate refs while we have c->erase_completion_lock held.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
We don't allocate this locally any more -- it's given to us and owner by
our caller. Also improve the debug messages a little.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Next step in ongoing campaign to file a struct jffs2_raw_node_ref for every
piece of dirty space in the system, so that __totlen can be killed off....
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
If __totlen is going away, we need to pass the length in separately.
Also stop callers from needlessly setting ref->next_phys to NULL,
since that's done for them... and since that'll also be going away soon.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Make sure we allocate a ref for any dirty space which exists between nodes
which we find in an eraseblock summary.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Steve French