Re: Fw: reiser4 git repo

From: Johannes Buchner
Date: Fri Nov 27 2009 - 22:10:24 EST


Hi,

I had some issues with the reiser4 patch series from -mm.
First of all, I'm not sure what release they are based on. I used
a8a8a669e from Linus' git repository. I applied all reiser4 patches
(except reiser4-disable of course) from
http://userweb.kernel.org/~akpm/mmotm/, which worked fine.

I think these patches are based on the reiser4-2.6.30 patch.
I also compared them to the reiser4-2.6.31 patch, which adds some
missing pieces ([1] reiser4-from-2.6.31-patch.patch).

Additionally, some recent commits demand some changes, especially
since current_is_pdflush and generic_sync_sb_inodes are not available
any more (d8a8559cd7a9cc).
[2] reiser4-pdflush-comments.patch only fixes some comments which still
mention pdflush.
In [3] reiser4-generic_sync_sb_inodes.patch I tried to
replace generic_sync_sb_inodes. I am not 100% sure that this is
correct, but that is what I understood from the commit diff+message of
d8a8559cd7a9cc.

The original patch
'reiser4-vfs-add-super_operationssync_inodes' introduced the call from
sync_inodes_sb to the super operation sync_inodes(sb, NULL). This NULL
pointer is dereferenced in reiser4_sync_inodes, which is fixed by
[4] reiser4-sync_inodes-null-dereference.patch.

All 4 additional patches are attached. I can only say that it works on
my machine, before I had compilation issues and freezes on 'sync'.

Cheers,
Johannes

PS: The git repo mentioned in
http://userweb.kernel.org/~akpm/mmotm/mmotm-readme.txt is not available.

On Fri, 27 Nov 2009 04:00:51 +0100
Frederic Weisbecker <fweisbec@xxxxxxxxx> wrote:

> On Fri, Nov 27, 2009 at 03:51:48PM +1300, Johannes Buchner wrote:
> > [ I sent this 2 days ago to linux-fsdevel@xxxxxxxxxxxxxxx, maybe
> > I'll get more responses in here. ]
> >
> > Hi!
> >
> > I would like to keep a git repo up to date with current kernel
> > development, with reiser4 patched in. I am having trouble finding a
> > repo that provides reiser4. Is there one?
> >
> > Some people may have noticed the last patch doesn't apply cleanly
> > anymore. I made a few lines of modifications in my local repo. Maybe
> > that repo would be interesting for others too*.
> >
> > I plan to catch up with changes from kernel development on a
> > daily/weekly basis; my /home partition is reiser4, so I am fairly
> > serious about it.
> >
> > At the moment, I fetch from torvalds/linux-2.6.git and
> > stable/linux-2.6.31.y.git. Am I missing any other repositories that
> > I should try to keep up with? (A fs subsystem maintainer repo?)
> > next/linux-next.git is probably moving too fast for me.
>
>
> You can try the -mm tree, the reiser 4 patches can be found inside:
>
> http://userweb.kernel.org/~akpm/mmotm/

--
Emails können geändert, gefälscht und eingesehen werden. Signiere oder
verschüssele deine Mails mit GPG.
http://web.student.tuwien.ac.at/~e0625457/pgp.html


Attachment: series
Description: Binary data

From: Johannes Buchner <buchner.johannes@xxxxxx>

some changes from reiser4-2.6.31 patch
from http://www.kernel.org/pub/linux/kernel/people/edward/reiser4/reiser4-for-2.6/

Index: linux/fs/reiser4/carry.c
===================================================================
--- linux.orig/fs/reiser4/carry.c
+++ linux/fs/reiser4/carry.c
@@ -1095,7 +1095,7 @@ static void fatal_carry_error(carry_leve
*
* This function itself only manages changes in carry structures and delegates
* all hard work (allocation of znode for new root, changes of parent and
- * sibling pointers to the reiser4_add_tree_root().
+ * sibling pointers) to the reiser4_add_tree_root().
*
* Locking: old tree root is locked by carry at this point. Fake znode is also
* locked.
Index: linux/fs/reiser4/carry_ops.c
===================================================================
--- linux.orig/fs/reiser4/carry_ops.c
+++ linux/fs/reiser4/carry_ops.c
@@ -2015,7 +2015,7 @@ static int carry_estimate_bitmaps(void)
int bytes;

bytes = capped_height() * (0 + /* bnode should be added, but
- * its is private to bitmap.c,
+ * it is private to bitmap.c,
* skip for now. */
2 * sizeof(jnode));
/* working and commit jnodes */
Index: linux/fs/reiser4/coord.c
===================================================================
--- linux.orig/fs/reiser4/coord.c
+++ linux/fs/reiser4/coord.c
@@ -36,12 +36,12 @@ void coord_normalize(coord_t *coord)

coord_clear_iplug(coord);

- if (node_is_empty(node))
+ if (node_is_empty(node)) {
coord_init_first_unit(coord, node);
- else if ((coord->between == AFTER_ITEM)
- || (coord->between == AFTER_UNIT))
+ } else if ((coord->between == AFTER_ITEM)
+ || (coord->between == AFTER_UNIT)) {
return;
- else if (coord->item_pos == coord_num_items(coord)
+ } else if (coord->item_pos == coord_num_items(coord)
&& coord->between == BEFORE_ITEM) {
coord_dec_item_pos(coord);
coord->between = AFTER_ITEM;
Index: linux/fs/reiser4/entd.c
===================================================================
--- linux.orig/fs/reiser4/entd.c
+++ linux/fs/reiser4/entd.c
@@ -241,7 +241,7 @@ static void entd_flush(struct super_bloc
if (rq->wbc->nr_to_write > 0) {
rq->wbc->range_start = 0;
rq->wbc->range_end = LLONG_MAX;
- generic_sync_sb_inodes(rq->wbc);
+ generic_sync_sb_inodes(super, rq->wbc);
}
rq->wbc->nr_to_write = ENTD_CAPTURE_APAGE_BURST;
reiser4_writeout(super, rq->wbc);
Index: linux/fs/reiser4/super_ops.c
===================================================================
--- linux.orig/fs/reiser4/super_ops.c
+++ linux/fs/reiser4/super_ops.c
@@ -412,7 +412,7 @@ static void reiser4_sync_inodes(struct s
* call reiser4_writepages for each of dirty inodes to turn dirty pages
* into transactions if they were not yet.
*/
- generic_sync_sb_inodes(wbc);
+ generic_sync_sb_inodes(super, wbc);

/* flush goes here */
wbc->nr_to_write = to_write;
Index: linux/fs/reiser4/znode.h
===================================================================
--- linux.orig/fs/reiser4/znode.h
+++ linux/fs/reiser4/znode.h
@@ -18,7 +18,6 @@

#include <linux/types.h>
#include <linux/spinlock.h>
-#include <linux/semaphore.h>
#include <linux/pagemap.h> /* for PAGE_CACHE_SIZE */
#include <asm/atomic.h>

Index: linux/include/linux/mm.h
===================================================================
--- linux.orig/include/linux/mm.h
+++ linux/include/linux/mm.h
@@ -836,6 +836,7 @@ int redirty_page_for_writepage(struct wr
void account_page_dirtied(struct page *page, struct address_space *mapping);
int set_page_dirty(struct page *page);
int set_page_dirty_lock(struct page *page);
+int set_page_dirty_notag(struct page *page);
int clear_page_dirty_for_io(struct page *page);

extern unsigned long move_page_tables(struct vm_area_struct *vma,
Index: linux/mm/filemap.c
===================================================================
--- linux.orig/mm/filemap.c
+++ linux/mm/filemap.c
@@ -139,6 +139,7 @@ void __remove_from_page_cache(struct pag
dec_bdi_stat(mapping->backing_dev_info, BDI_RECLAIMABLE);
}
}
+EXPORT_SYMBOL(__remove_from_page_cache);

void remove_from_page_cache(struct page *page)
{
@@ -151,6 +152,7 @@ void remove_from_page_cache(struct page
spin_unlock_irq(&mapping->tree_lock);
mem_cgroup_uncharge_cache_page(page);
}
+EXPORT_SYMBOL(remove_from_page_cache);

static int sync_page(void *word)
{
@@ -800,7 +802,6 @@ repeat:
rcu_read_unlock();
return ret;
}
-EXPORT_SYMBOL(find_get_pages);

/**
* find_get_pages_contig - gang contiguous pagecache lookup
@@ -970,6 +971,7 @@ static void shrink_readahead_size_eio(st
{
ra->ra_pages /= 4;
}
+EXPORT_SYMBOL(find_get_pages);

/**
* do_generic_file_read - generic file read routine
Index: linux/mm/page-writeback.c
===================================================================
--- linux.orig/mm/page-writeback.c
+++ linux/mm/page-writeback.c
@@ -1143,6 +1143,32 @@ int __set_page_dirty_nobuffers(struct pa
EXPORT_SYMBOL(__set_page_dirty_nobuffers);

/*
+ * set_page_dirty_notag() -- similar to __set_page_dirty_nobuffers()
+ * except it doesn't tag the page dirty in the page-cache radix tree.
+ * This means that the address space using this cannot use the regular
+ * filemap ->writepages() helpers and must provide its own means of
+ * tracking and finding non-tagged dirty pages.
+ *
+ * NOTE: furthermore, this version also doesn't handle truncate races.
+ */
+int set_page_dirty_notag(struct page *page)
+{
+ struct address_space *mapping = page->mapping;
+
+ if (!TestSetPageDirty(page)) {
+ unsigned long flags;
+ WARN_ON_ONCE(!PagePrivate(page) && !PageUptodate(page));
+ local_irq_save(flags);
+ account_page_dirtied(page, mapping);
+ local_irq_restore(flags);
+ __mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
+ return 1;
+ }
+ return 0;
+}
+EXPORT_SYMBOL(set_page_dirty_notag);
+
+/*
* When a writepage implementation decides that it doesn't want to write this
* page for some reason, it should redirty the locked page via
* redirty_page_for_writepage() and it should then unlock the page and return 0

From: Johannes Buchner <buchner.johannes@xxxxxx>

some comments were still mentioning pdflush

diff --git a/fs/reiser4/context.c b/fs/reiser4/context.c
index eed3efa..f80ae13 100644
--- a/fs/reiser4/context.c
+++ b/fs/reiser4/context.c
@@ -147,7 +147,7 @@ static void reiser4_throttle_write_at(reiser4_context *context)
* call balance_dirty_pages_ratelimited() to process formatted nodes
* dirtied during this system call. Do that only if we are not in mount
* and there were nodes dirtied in this context and we are not in
- * writepage (to avoid deadlock) and not in pdflush
+ * writepage (to avoid deadlock)
*/
if (sbinfo != NULL && sbinfo->fake != NULL &&
context->nr_marked_dirty != 0 &&
diff --git a/fs/reiser4/txnmgr.c b/fs/reiser4/txnmgr.c
index 74b49f2..844d19e 100644
--- a/fs/reiser4/txnmgr.c
+++ b/fs/reiser4/txnmgr.c
@@ -1360,7 +1360,7 @@ static int txn_try_to_fuse_small_atom(txn_mgr * tmgr, txn_atom * atom)
code tries to flush current atom.

flush_some_atom() is called as part of memory clearing process. It is
- invoked from balance_dirty_pages(), pdflushd, and entd.
+ invoked from balance_dirty_pages() and entd.

If we can flush no nodes, atom is committed, because this frees memory.

@@ -1469,7 +1469,7 @@ flush_some_atom(jnode * start, long *nr_submitted, const struct writeback_contro
* or atom is too old/large,
* we force current atom to commit */
/* wait for commit completion but only if this
- * wouldn't stall pdflushd and ent thread. */
+ * wouldn't stall ent thread. */
if (!wbc->nonblocking && !ctx->entd)
txnh->flags |= TXNH_WAIT_COMMIT;
atom->flags |= ATOM_FORCE_COMMIT;

From: Johannes Buchner <buchner.johannes@xxxxxx>

generic_sync_sb_inodes doesn't exist anymore

Index: linux/fs/reiser4/entd.c
===================================================================
--- linux.orig/fs/reiser4/entd.c
+++ linux/fs/reiser4/entd.c
@@ -241,7 +241,9 @@ static void entd_flush(struct super_bloc
if (rq->wbc->nr_to_write > 0) {
rq->wbc->range_start = 0;
rq->wbc->range_end = LLONG_MAX;
- generic_sync_sb_inodes(super, rq->wbc);
+ writeback_inodes_sb(super);
+ if (rq->wbc->sync_mode == WB_SYNC_ALL)
+ sync_inodes_sb(super);
}
rq->wbc->nr_to_write = ENTD_CAPTURE_APAGE_BURST;
reiser4_writeout(super, rq->wbc);
Index: linux/fs/reiser4/super_ops.c
===================================================================
--- linux.orig/fs/reiser4/super_ops.c
+++ linux/fs/reiser4/super_ops.c
@@ -412,7 +412,9 @@ static void reiser4_sync_inodes(struct s
* call reiser4_writepages for each of dirty inodes to turn dirty pages
* into transactions if they were not yet.
*/
- generic_sync_sb_inodes(super, wbc);
+ writeback_inodes_sb(super);
+ if (wbc->sync_mode == WB_SYNC_ALL)
+ sync_inodes_sb(super);

/* flush goes here */
wbc->nr_to_write = to_write;

From: Johannes Buchner <buchner.johannes@xxxxxx>

the parameter wbc=NULL caused a NULL pointer dereference in
reiser4_sync_inodes.

introduced in patch reiser4-vfs-add-super_operationssync_inodes
Index: linux/fs/reiser4/super_ops.c
===================================================================
--- linux.orig/fs/reiser4/super_ops.c
+++ linux/fs/reiser4/super_ops.c
@@ -395,7 +395,7 @@ static void reiser4_sync_inodes(struct s
reiser4_context *ctx;
long to_write;

- if (wbc->for_kupdate)
+ if (wbc == NULL || wbc->for_kupdate)
/* reiser4 has its own means of periodical write-out */
return;


Attachment: pgp00000.pgp
Description: PGP signature