---

yaml
---
r: 109729
b: refs/heads/master
c: 3d6e48f
h: refs/heads/master
i:
  109727: e8184dd
v: v3

[refs]

@@ -1,2 +1,2 @@
 ---
-refs/heads/master: 0b1fc335d2f2c1206ac4048e5f6d8971f2aae6be
+refs/heads/master: 3d6e48f43340343d97839eadb1ab7b6a3ea98797

trunk/Documentation/filesystems/Locking

@@ -144,8 +144,8 @@ prototypes:
 	void (*kill_sb) (struct super_block *);
 locking rules:
 		may block	BKL
-get_sb		yes		no
-kill_sb		yes		no
+get_sb		yes		yes
+kill_sb		yes		yes
 
 ->get_sb() returns error or 0 with locked superblock attached to the vfsmount
 (exclusive on ->s_umount).
@@ -409,12 +409,12 @@ ioctl:			yes	(see below)
 unlocked_ioctl:		no	(see below)
 compat_ioctl:		no
 mmap:			no
-open:			no
+open:			maybe	(see below)
 flush:			no
 release:		no
 fsync:			no	(see below)
 aio_fsync:		no
-fasync:			no
+fasync:			yes	(see below)
 lock:			yes
 readv:			no
 writev:			no
@@ -431,6 +431,13 @@ For many filesystems, it is probably safe to acquire the inode
 semaphore.  Note some filesystems (i.e. remote ones) provide no
 protection for i_size so you will need to use the BKL.
 
+->open() locking is in-transit: big lock partially moved into the methods.
+The only exception is ->open() in the instances of file_operations that never
+end up in ->i_fop/->proc_fops, i.e. ones that belong to character devices
+(chrdev_open() takes lock before replacing ->f_op and calling the secondary
+method. As soon as we fix the handling of module reference counters all
+instances of ->open() will be called without the BKL.
+
 Note: ext2_release() was *the* source of contention on fs-intensive
 loads and dropping BKL on ->release() helps to get rid of that (we still
 grab BKL for cases when we close a file that had been opened r/w, but that

trunk/MAINTAINERS

@@ -750,13 +750,11 @@ P:	Ville Syrjala
 M:	syrjala@sci.fi
 S:	Maintained
 
-ATLX ETHERNET DRIVERS
+ATL1 ETHERNET DRIVER
 P:	Jay Cliburn
 M:	jcliburn@gmail.com
 P:	Chris Snook
 M:	csnook@redhat.com
-P:	Jie Yang
-M:	jie.yang@atheros.com
 L:	atl1-devel@lists.sourceforge.net
 W:	http://sourceforge.net/projects/atl1
 W:	http://atl1.sourceforge.net

trunk/arch/ia64/include/asm/sections.h

@@ -21,8 +21,5 @@ extern char __start_gate_brl_fsys_bubble_down_patchlist[], __end_gate_brl_fsys_b
 extern char __start_unwind[], __end_unwind[];
 extern char __start_ivt_text[], __end_ivt_text[];
 
-#undef dereference_function_descriptor
-void *dereference_function_descriptor(void *);
-
 #endif /* _ASM_IA64_SECTIONS_H */
 

trunk/arch/ia64/kernel/module.c

@@ -31,11 +31,9 @@
 #include <linux/elf.h>
 #include <linux/moduleloader.h>
 #include <linux/string.h>
-#include <linux/uaccess.h>
 #include <linux/vmalloc.h>
 
 #include <asm/patch.h>
-#include <asm/sections.h>
 #include <asm/unaligned.h>
 
 #define ARCH_MODULE_DEBUG 0
@@ -943,13 +941,3 @@ module_arch_cleanup (struct module *mod)
 	if (mod->arch.core_unw_table)
 		unw_remove_unwind_table(mod->arch.core_unw_table);
 }
-
-void *dereference_function_descriptor(void *ptr)
-{
-	struct fdesc *desc = ptr;
-	void *p;
-
-	if (!probe_kernel_address(&desc->ip, p))
-		ptr = p;
-	return ptr;
-}

trunk/arch/mips/sgi-ip22/ip22-platform.c

@@ -150,7 +150,7 @@ static int __init sgiseeq_devinit(void)
 		return res;
 
 	/* Second HPC is missing? */
-	if (ip22_is_fullhouse() ||
+	if (!ip22_is_fullhouse() ||
 	    get_dbe(tmp, (unsigned int *)&hpc3c1->pbdma[1]))
 		return 0;
 

trunk/arch/parisc/kernel/module.c

@@ -47,9 +47,7 @@
 #include <linux/string.h>
 #include <linux/kernel.h>
 #include <linux/bug.h>
-#include <linux/uaccess.h>
 
-#include <asm/sections.h>
 #include <asm/unwind.h>
 
 #if 0
@@ -862,15 +860,3 @@ void module_arch_cleanup(struct module *mod)
 	deregister_unwind_table(mod);
 	module_bug_cleanup(mod);
 }
-
-#ifdef CONFIG_64BIT
-void *dereference_function_descriptor(void *ptr)
-{
-	Elf64_Fdesc *desc = ptr;
-	void *p;
-
-	if (!probe_kernel_address(&desc->addr, p))
-		ptr = p;
-	return ptr;
-}
-#endif

trunk/arch/powerpc/include/asm/sections.h

@@ -16,9 +16,6 @@ static inline int in_kernel_text(unsigned long addr)
 	return 0;
 }
 
-#undef dereference_function_descriptor
-void *dereference_function_descriptor(void *);
-
 #endif
 
 #endif /* __KERNEL__ */

trunk/arch/powerpc/kernel/module_64.c

@@ -21,9 +21,8 @@
 #include <linux/err.h>
 #include <linux/vmalloc.h>
 #include <linux/bug.h>
-#include <linux/uaccess.h>
 #include <asm/module.h>
-#include <asm/sections.h>
+#include <asm/uaccess.h>
 #include <asm/firmware.h>
 #include <asm/code-patching.h>
 #include <linux/sort.h>
@@ -452,13 +451,3 @@ int apply_relocate_add(Elf64_Shdr *sechdrs,
 
 	return 0;
 }
-
-void *dereference_function_descriptor(void *ptr)
-{
-	struct ppc64_opd_entry *desc = ptr;
-	void *p;
-
-	if (!probe_kernel_address(&desc->funcaddr, p))
-		ptr = p;
-	return ptr;
-}

trunk/arch/s390/kernel/compat_ptrace.h

@@ -42,6 +42,7 @@ struct user_regs_struct32
 	u32 gprs[NUM_GPRS];
 	u32 acrs[NUM_ACRS];
 	u32 orig_gpr2;
+	/* nb: there's a 4-byte hole here */
 	s390_fp_regs fp_regs;
 	/*
 	 * These per registers are in here so that gdb can modify them

trunk/arch/s390/kernel/ptrace.c

@@ -170,6 +170,13 @@ static unsigned long __peek_user(struct task_struct *child, addr_t addr)
 		 */
 		tmp = (addr_t) task_pt_regs(child)->orig_gpr2;
 
+	} else if (addr < (addr_t) &dummy->regs.fp_regs) {
+		/*
+		 * prevent reads of padding hole between
+		 * orig_gpr2 and fp_regs on s390.
+		 */
+		tmp = 0;
+
 	} else if (addr < (addr_t) (&dummy->regs.fp_regs + 1)) {
 		/* 
 		 * floating point regs. are stored in the thread structure
@@ -270,6 +277,13 @@ static int __poke_user(struct task_struct *child, addr_t addr, addr_t data)
 		 */
 		task_pt_regs(child)->orig_gpr2 = data;
 
+	} else if (addr < (addr_t) &dummy->regs.fp_regs) {
+		/*
+		 * prevent writes of padding hole between
+		 * orig_gpr2 and fp_regs on s390.
+		 */
+		return 0;
+
 	} else if (addr < (addr_t) (&dummy->regs.fp_regs + 1)) {
 		/*
 		 * floating point regs. are stored in the thread structure
@@ -428,6 +442,13 @@ static u32 __peek_user_compat(struct task_struct *child, addr_t addr)
 		 */
 		tmp = *(__u32*)((addr_t) &task_pt_regs(child)->orig_gpr2 + 4);
 
+	} else if (addr < (addr_t) &dummy32->regs.fp_regs) {
+		/*
+		 * prevent reads of padding hole between
+		 * orig_gpr2 and fp_regs on s390.
+		 */
+		tmp = 0;
+
 	} else if (addr < (addr_t) (&dummy32->regs.fp_regs + 1)) {
 		/*
 		 * floating point regs. are stored in the thread structure 
@@ -514,6 +535,13 @@ static int __poke_user_compat(struct task_struct *child,
 		 */
 		*(__u32*)((addr_t) &task_pt_regs(child)->orig_gpr2 + 4) = tmp;
 
+	} else if (addr < (addr_t) &dummy32->regs.fp_regs) {
+		/*
+		 * prevent writess of padding hole between
+		 * orig_gpr2 and fp_regs on s390.
+		 */
+		return 0;
+
 	} else if (addr < (addr_t) (&dummy32->regs.fp_regs + 1)) {
 		/*
 		 * floating point regs. are stored in the thread structure 

trunk/fs/ubifs/budget.c

@@ -302,6 +302,18 @@ long long ubifs_calc_available(const struct ubifs_info *c, int min_idx_lebs)
 	int subtract_lebs;
 	long long available;
 
+	/*
+	 * Force the amount available to the total size reported if the used
+	 * space is zero.
+	 */
+	if (c->lst.total_used <= UBIFS_INO_NODE_SZ &&
+	    c->budg_data_growth + c->budg_dd_growth == 0) {
+		/* Do the same calculation as for c->block_cnt */
+		available = c->main_lebs - 2;
+		available *= c->leb_size - c->dark_wm;
+		return available;
+	}
+
 	available = c->main_bytes - c->lst.total_used;
 
 	/*
@@ -702,106 +714,34 @@ void ubifs_release_dirty_inode_budget(struct ubifs_info *c,
 }
 
 /**
- * ubifs_reported_space - calculate reported free space.
- * @c: the UBIFS file-system description object
- * @free: amount of free space
- *
- * This function calculates amount of free space which will be reported to
- * user-space. User-space application tend to expect that if the file-system
- * (e.g., via the 'statfs()' call) reports that it has N bytes available, they
- * are able to write a file of size N. UBIFS attaches node headers to each data
- * node and it has to write indexind nodes as well. This introduces additional
- * overhead, and UBIFS it has to report sligtly less free space to meet the
- * above expectetion.
- *
- * This function assumes free space is made up of uncompressed data nodes and
- * full index nodes (one per data node, tripled because we always allow enough
- * space to write the index thrice).
- *
- * Note, the calculation is pessimistic, which means that most of the time
- * UBIFS reports less space than it actually has.
- */
-long long ubifs_reported_space(const struct ubifs_info *c, uint64_t free)
-{
-	int divisor, factor, f;
-
-	/*
-	 * Reported space size is @free * X, where X is UBIFS block size
-	 * divided by UBIFS block size + all overhead one data block
-	 * introduces. The overhead is the node header + indexing overhead.
-	 *
-	 * Indexing overhead calculations are based on the following formula:
-	 * I = N/(f - 1) + 1, where I - number of indexing nodes, N - number
-	 * of data nodes, f - fanout. Because effective UBIFS fanout is twice
-	 * as less than maximum fanout, we assume that each data node
-	 * introduces 3 * @c->max_idx_node_sz / (@c->fanout/2 - 1) bytes.
-	 * Note, the multiplier 3 is because UBIFS reseves thrice as more space
-	 * for the index.
-	 */
-	f = c->fanout > 3 ? c->fanout >> 1 : 2;
-	factor = UBIFS_BLOCK_SIZE;
-	divisor = UBIFS_MAX_DATA_NODE_SZ;
-	divisor += (c->max_idx_node_sz * 3) / (f - 1);
-	free *= factor;
-	do_div(free, divisor);
-	return free;
-}
-
-/**
- * ubifs_get_free_space - return amount of free space.
+ * ubifs_budg_get_free_space - return amount of free space.
  * @c: UBIFS file-system description object
  *
- * This function calculates amount of free space to report to user-space.
- *
- * Because UBIFS may introduce substantial overhead (the index, node headers,
- * alighment, wastage at the end of eraseblocks, etc), it cannot report real
- * amount of free flash space it has (well, because not all dirty space is
- * reclamable, UBIFS does not actually know the real amount). If UBIFS did so,
- * it would bread user expectetion about what free space is. Users seem to
- * accustomed to assume that if the file-system reports N bytes of free space,
- * they would be able to fit a file of N bytes to the FS. This almost works for
- * traditional file-systems, because they have way less overhead than UBIFS.
- * So, to keep users happy, UBIFS tries to take the overhead into account.
+ * This function returns amount of free space on the file-system.
  */
-long long ubifs_get_free_space(struct ubifs_info *c)
+long long ubifs_budg_get_free_space(struct ubifs_info *c)
 {
-	int min_idx_lebs, rsvd_idx_lebs, lebs;
+	int min_idx_lebs, rsvd_idx_lebs;
 	long long available, outstanding, free;
 
+	/* Do exactly the same calculations as in 'do_budget_space()' */
 	spin_lock(&c->space_lock);
 	min_idx_lebs = ubifs_calc_min_idx_lebs(c);
-	outstanding = c->budg_data_growth + c->budg_dd_growth;
 
-	/*
-	 * Force the amount available to the total size reported if the used
-	 * space is zero.
-	 */
-	if (c->lst.total_used <= UBIFS_INO_NODE_SZ && !outstanding) {
-		spin_unlock(&c->space_lock);
-		return (long long)c->block_cnt << UBIFS_BLOCK_SHIFT;
-	}
-
-	available = ubifs_calc_available(c, min_idx_lebs);
-
-	/*
-	 * When reporting free space to user-space, UBIFS guarantees that it is
-	 * possible to write a file of free space size. This means that for
-	 * empty LEBs we may use more precise calculations than
-	 * 'ubifs_calc_available()' is using. Namely, we know that in empty
-	 * LEBs we would waste only @c->leb_overhead bytes, not @c->dark_wm.
-	 * Thus, amend the available space.
-	 *
-	 * Note, the calculations below are similar to what we have in
-	 * 'do_budget_space()', so refer there for comments.
-	 */
 	if (min_idx_lebs > c->lst.idx_lebs)
 		rsvd_idx_lebs = min_idx_lebs - c->lst.idx_lebs;
 	else
 		rsvd_idx_lebs = 0;
-	lebs = c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt -
-	       c->lst.taken_empty_lebs;
-	lebs -= rsvd_idx_lebs;
-	available += lebs * (c->dark_wm - c->leb_overhead);
+
+	if (rsvd_idx_lebs > c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt
+				- c->lst.taken_empty_lebs) {
+		spin_unlock(&c->space_lock);
+		return 0;
+	}
+
+	available = ubifs_calc_available(c, min_idx_lebs);
+	outstanding = c->budg_data_growth + c->budg_dd_growth;
+	c->min_idx_lebs = min_idx_lebs;
 	spin_unlock(&c->space_lock);
 
 	if (available > outstanding)

trunk/fs/ubifs/dir.c

@@ -587,6 +587,7 @@ static int ubifs_unlink(struct inode *dir, struct dentry *dentry)
 	if (err) {
 		if (err != -ENOSPC)
 			return err;
+		err = 0;
 		budgeted = 0;
 	}