NOMMU: Provide shared-writable mmap support on ramfs
Subsystems (4)
PATCH 1/3 NOMMU: Provide shared-writable mmap support on ramfs
The attached patch makes ramfs support shared-writable mmaps by:
(1) Attempting to perform a contiguous block allocation to the requested size
when truncate attempts to increase the file from zero size, such as
happens when:
fd = shm_open("/file/on/ramfs", ...):
ftruncate(fd, size_requested);
addr = mmap(NULL, subsize, PROT_READ|PROT_WRITE|PROT_EXEC, MAP_SHARED,
fd, offset); (2) Permitting any shared-writable mapping over any contiguous set of extant pages. get_unmapped_area() will return the address into the actual ramfs pages. The mapping may start anywhere and be of any size, but may not go over the end of file. Multiple mappings may overlap in any way.
(3) Not permitting a file to be shrunk if it would truncate any shared mappings (private mappings are copied).
Thus this patch provides support for POSIX shared memory on NOMMU kernels, with certain limitations such as there being a large enough block of pages available to support the allocation and it only working on directly mappable filesystems.
---
warthog>diffstat -p1 shmem-ramfs-2615rc2.diff
fs/ramfs/Makefile | 4 fs/ramfs/file-mmu.c | 57 +++++++++ fs/ramfs/file-nommu.c | 294 ++++++++++++++++++++++++++++++++++++++++++++++++++ fs/ramfs/inode.c | 22 --- fs/ramfs/internal.h | 15 ++ include/linux/ramfs.h | 10 + 6 files changed, 380 insertions(+), 22 deletions(-)
fs/ramfs/file-mmu.c +57 -0
@@ -0,0 +1,57 @@ +/* file-mmu.c: ramfs MMU-based file operations +* +* Resizable simple ram filesystem for Linux. +* +* Copyright (C) 2000 Linus Torvalds. +* 2000 Transmeta Corp. +* +* Usage limits added by David Gibson, Linuxcare Australia. +* This file is released under the GPL. +*/ + +/* +* NOTE! This filesystem is probably most useful +* not as a real filesystem, but as an example of +* how virtual filesystems can be written. +* +* It doesn't get much simpler than this. Consider +* that this file implements the full semantics of +* a POSIX-compliant read-write filesystem. +* +* Note in particular how the filesystem does not +* need to implement any data structures of its own +* to keep track of the virtual data: using the VFS +* caches is sufficient. +*/ + +#include <linux/module.h> +#include <linux/fs.h> +#include <linux/pagemap.h> +#include <linux/highmem.h> +#include <linux/init.h> +#include <linux/string.h> +#include <linux/smp_lock.h> +#include <linux/backing-dev.h> +#include <linux/ramfs.h> + +#include <asm/uaccess.h> +#include "internal.h" + +struct address_space_operations ramfs_aops = { + .readpage = simple_readpage, + .prepare_write = simple_prepare_write, + .commit_write = simple_commit_write +}; + +struct file_operations ramfs_file_operations = { + .read = generic_file_read, + .write = generic_file_write, + .mmap = generic_file_mmap, + .fsync = simple_sync_file, + .sendfile = generic_file_sendfile, + .llseek = generic_file_llseek, +}; + +struct inode_operations ramfs_file_inode_operations = { + .getattr = simple_getattr, +};
fs/ramfs/file-nommu.c +294 -0
294 lines not on this page · raw message
fs/ramfs/inode.c +1 -21
@@ -34,13 +34,12 @@#include <linux/ramfs.h> #include <asm/uaccess.h> +#include "internal.h" /* some random number */ #define RAMFS_MAGIC 0x858458f6static struct super_operations ramfs_ops; -static struct address_space_operations ramfs_aops; -static struct inode_operations ramfs_file_inode_operations;static struct inode_operations ramfs_dir_inode_operations;static struct backing_dev_info ramfs_backing_dev_info = {
@@ -142,25 +141,6 @@ static int ramfs_symlink(struct inode *return error; } -static struct address_space_operations ramfs_aops = { - .readpage = simple_readpage, - .prepare_write = simple_prepare_write, - .commit_write = simple_commit_write -}; - -struct file_operations ramfs_file_operations = { - .read = generic_file_read, - .write = generic_file_write, - .mmap = generic_file_mmap, - .fsync = simple_sync_file, - .sendfile = generic_file_sendfile, - .llseek = generic_file_llseek, -}; - -static struct inode_operations ramfs_file_inode_operations = { - .getattr = simple_getattr, -}; -static struct inode_operations ramfs_dir_inode_operations = { .create = ramfs_create, .lookup = simple_lookup,
fs/ramfs/internal.h +15 -0
@@ -0,0 +1,15 @@ +/* internal.h: ramfs internal definitions +* +* Copyright (C) 2005 Red Hat, Inc. All Rights Reserved. +* Written by David Howells ([email protected]) +* +* This program is free software; you can redistribute it and/or +* modify it under the terms of the GNU General Public License +* as published by the Free Software Foundation; either version +* 2 of the License, or (at your option) any later version. +*/ + + +extern struct address_space_operations ramfs_aops; +extern struct file_operations ramfs_file_operations; +extern struct inode_operations ramfs_file_inode_operations;
fs/ramfs/Makefile +3 -1
@@ -4,4 +4,6 @@ obj-$(CONFIG_RAMFS) += ramfs.o -ramfs-objs := inode.o +file-mmu-y := file-nommu.o +file-mmu-$(CONFIG_MMU) := file-mmu.o +ramfs-objs += inode.o $(file-mmu-y)
include/linux/ramfs.h +10 -0
@@ -5,6 +5,16 @@ struct inode *ramfs_get_inode(struct supstruct super_block *ramfs_get_sb (struct file_system_type *fs_type,int flags,const char *dev_name,void *data); +#ifndef CONFIG_MMU +extern unsigned long ramfs_nommu_get_unmapped_area (struct file *file, +unsigned long addr, +unsigned long len, +unsigned long pgoff, +unsigned long flags); + +extern int ramfs_nommu_mmap (struct file *file,struct vm_area_struct *vma); +#endif +extern struct file_operations ramfs_file_operations;extern struct vm_operations_struct generic_file_vm_ops;
PATCH 2/3 NOMMU: Make SYSV IPC SHM use ramfs facilities on NOMMU
The attached patch makes the SYSV IPC shared memory facilities use the new ramfs facilities on a no-MMU kernel.
The following changes are made:
(1) There are now shmem_mmap() and shmem_get_unmapped_area() functions to allow the IPC SHM facilities to commune with the tiny-shmem and shmem code.
(2) ramfs files now need resizing using do_truncate() rather than by modifying the inode size directly (see shmem_file_setup()). This causes ramfs to attempt to bind a block of pages of sufficient size to the inode.
(3) CONFIG_SYSVIPC is no longer contingent on CONFIG_MMU.
---
warthog>diffstat -p1 shmem-nommu-2615rc2.diff
include/linux/mm.h | 9 +++++++++ init/Kconfig | 1 - ipc/shm.c | 18 +++++++++++++----- mm/nommu.c | 6 ++++++ mm/shmem.c | 2 +- mm/tiny-shmem.c | 29 ++++++++++++++++++++++++++++- 6 files changed, 57 insertions(+), 8 deletions(-)
include/linux/mm.h +9 -0
@@ -656,9 +656,18 @@ int shmem_lock(struct file *file, int lo#define shmem_get_policy (a, b) (NULL)#endif struct file *shmem_file_setup (char *name,loff_t size,unsigned long flags); +extern int shmem_mmap (struct file *file,struct vm_area_struct *vma);int shmem_zero_setup (struct vm_area_struct *); +#ifndef CONFIG_MMU +extern unsigned long shmem_get_unmapped_area (struct file *file, +unsigned long addr, +unsigned long len, +unsigned long pgoff, +unsigned long flags); +#endif +static inline int can_do_mlock (void ) {if (capable (CAP_IPC_LOCK ))
init/Kconfig +0 -1
@@ -105,7 +105,6 @@ config SWAP config SYSVIPC bool "System V IPC" - depends on MMU ---help--- Inter Process Communication is a suite of library functions and system calls which let processes (running programs) synchronize and
ipc/shm.c +13 -5
@@ -157,14 +157,22 @@ static void shm_close (struct vm_area_ststatic int shm_mmap (struct file * file,struct vm_area_struct * vma) { -file_accessed (file); - vma->vm_ops = &shm_vm_ops; -shm_inc (file->f_dentry->d_inode->i_ino); -return 0 ; +int ret; + + ret =shmem_mmap (file, vma); +if (ret ==0 ) { + vma->vm_ops = &shm_vm_ops; +shm_inc (file->f_dentry->d_inode->i_ino); + } + +return ret; }static struct file_operations shm_file_operations = { - .mmap = shm_mmap + .mmap = shm_mmap, +#ifndef CONFIG_MMU + .get_unmapped_area = shmem_get_unmapped_area, +#endif };static struct vm_operations_struct shm_vm_ops = {
mm/nommu.c +6 -0
@@ -1177,3 +1177,9 @@ int in_gate_area_no_task(unsigned long a {return 0 ; } + +struct page *filemap_nopage (struct vm_area_struct * area,unsigned long address,int *type) +{ +BUG (); +return NULL ; +}
mm/shmem.c +1 -1
@@ -1255,7 +1255,7 @@ out_nomem:return retval; } -static int shmem_mmap (struct file *file,struct vm_area_struct *vma) +int shmem_mmap (struct file *file,struct vm_area_struct *vma) {file_accessed (file); vma->vm_ops = &shmem_vm_ops;
mm/tiny-shmem.c +28 -1
@@ -81,13 +81,19 @@ struct file *shmem_file_setup(char *name goto close_file;d_instantiate (dentry, inode); - inode->i_size = size; inode->i_nlink =0 ;/* It is unlinked */ + file->f_vfsmnt =mntget (shm_mnt); file->f_dentry = dentry; file->f_mapping = inode->i_mapping; file->f_op = &ramfs_file_operations; file->f_mode =FMODE_WRITE |FMODE_READ ; + +/* notify everyone as to the change of file size */ + error =do_truncate (dentry, size, file); +if (error <0 ) + goto close_file; +return file; close_file:
@@ -123,3 +129,24 @@ int shmem_unuse(swp_entry_t entry, struc {return 0 ; } + +int shmem_mmap (struct file *file,struct vm_area_struct *vma) +{ +file_accessed (file); +#ifndef CONFIG_MMU +return ramfs_nommu_mmap (file, vma); +#else +return 0 ; +#endif +} + +#ifndef CONFIG_MMU +unsigned long shmem_get_unmapped_area (struct file *file, +unsigned long addr, +unsigned long len, +unsigned long pgoff, +unsigned long flags) +{ +return ramfs_nommu_get_unmapped_area (file, addr, len, pgoff, flags); +} +#endif
PATCH 3/3 FRV: Implement futex operations for FRV
In reply to David HowellsThe attached patch implements futex operations for the FRV architecture. The operations are applicable to both MMU and no-MMU modes; though the EFAULT handling will be a little bit of wasted space on the latter.
---
warthog>diffstat -p1 frv-futexop-2615rc2.diff
arch/frv/kernel/Makefile | 1 arch/frv/kernel/futex.c | 242 +++++++++++++++++++++++++++++++++++++++++++++++ include/asm-frv/futex.h | 42 -------- 3 files changed, 244 insertions(+), 41 deletions(-)
arch/frv/kernel/futex.c +242 -0
@@ -0,0 +1,242 @@ +/* futex.c: futex operations +* +* Copyright (C) 2005 Red Hat, Inc. All Rights Reserved. +* Written by David Howells ([email protected]) +* +* This program is free software; you can redistribute it and/or +* modify it under the terms of the GNU General Public License +* as published by the Free Software Foundation; either version +* 2 of the License, or (at your option) any later version. +*/ + +#include <linux/futex.h> +#include <asm/futex.h> +#include <asm/errno.h> +#include <asm/uaccess.h> + +/* +* the various futex operations; MMU fault checking is ignored under no-MMU +* conditions +*/ +static inline int atomic_futex_op_xchg_set (int oparg,int __user *uaddr,int *_oldval) +{ +int oldval, ret; + + asm("0: \n" +" orcc gr0,gr0,gr0,icc3 \n" /* set ICC3.Z */ +" ckeq icc3,cc7 \n" +"1: ld.p %M0,%1 \n" /* LD.P/ORCR must be atomic */ +" orcr cc7,cc7,cc3 \n" /* set CC3 to true */ +"2: cst.p %3,%M0 ,cc3,#1 \n" +" corcc gr29,gr29,gr0 ,cc3,#1 \n" /* clear ICC3.Z if store happens */ +" beq icc3,#0,0b \n" +" setlos 0,%2 \n" +"3: \n" +".subsection 2 \n" +"4: setlos %5,%2 \n" +" bra 3b \n" +".previous \n" +".section __ex_table,\"a\" \n" +" .balign 8 \n" +" .long 1b,4b \n" +" .long 2b,4b \n" +".previous" + :"+U" (*uaddr),"=&r" (oldval),"=&r" (ret),"=r" (oparg) + :"3" (oparg),"i" (-EFAULT ) + :"memory" ,"cc7" ,"cc3" ,"icc3" + ); + + *_oldval = oldval; +return ret; +} + +static inline int atomic_futex_op_xchg_add (int oparg,int __user *uaddr,int *_oldval) +{ +int oldval, ret; + + asm("0: \n" +" orcc gr0,gr0,gr0,icc3 \n" /* set ICC3.Z */ +" ckeq icc3,cc7 \n" +"1: ld.p %M0,%1 \n" /* LD.P/ORCR must be atomic */ +" orcr cc7,cc7,cc3 \n" /* set CC3 to true */ +" add %1,%3,%3 \n" +"2: cst.p %3,%M0 ,cc3,#1 \n" +" corcc gr29,gr29,gr0 ,cc3,#1 \n" /* clear ICC3.Z if store happens */ +" beq icc3,#0,0b \n" +" setlos 0,%2 \n" +"3: \n" +".subsection 2 \n" +"4: setlos %5,%2 \n" +" bra 3b \n" +".previous \n" +".section __ex_table,\"a\" \n" +" .balign 8 \n" +" .long 1b,4b \n" +" .long 2b,4b \n" +".previous" + :"+U" (*uaddr),"=&r" (oldval),"=&r" (ret),"=r" (oparg) + :"3" (oparg),"i" (-EFAULT ) + :"memory" ,"cc7" ,"cc3" ,"icc3" + ); + + *_oldval = oldval; +return ret; +} + +static inline int atomic_futex_op_xchg_or (int oparg,int __user *uaddr,int *_oldval) +{ +int oldval, ret; + + asm("0: \n" +" orcc gr0,gr0,gr0,icc3 \n" /* set ICC3.Z */ +" ckeq icc3,cc7 \n" +"1: ld.p %M0,%1 \n" /* LD.P/ORCR must be atomic */ +" orcr cc7,cc7,cc3 \n" /* set CC3 to true */ +" or %1,%3,%3 \n" +"2: cst.p %3,%M0 ,cc3,#1 \n" +" corcc gr29,gr29,gr0 ,cc3,#1 \n" /* clear ICC3.Z if store happens */ +" beq icc3,#0,0b \n" +" setlos 0,%2 \n" +"3: \n" +".subsection 2 \n" +"4: setlos %5,%2 \n" +" bra 3b \n" +".previous \n" +".section __ex_table,\"a\" \n" +" .balign 8 \n" +" .long 1b,4b \n" +" .long 2b,4b \n" +".previous" + :"+U" (*uaddr),"=&r" (oldval),"=&r" (ret),"=r" (oparg) + :"3" (oparg),"i" (-EFAULT ) + :"memory" ,"cc7" ,"cc3" ,"icc3" + ); + + *_oldval = oldval; +return ret; +} + +static inline int atomic_futex_op_xchg_and (int oparg,int __user *uaddr,int *_oldval) +{ +int oldval, ret; + + asm("0: \n" +" orcc gr0,gr0,gr0,icc3 \n" /* set ICC3.Z */ +" ckeq icc3,cc7 \n" +"1: ld.p %M0,%1 \n" /* LD.P/ORCR must be atomic */ +" orcr cc7,cc7,cc3 \n" /* set CC3 to true */ +" and %1,%3,%3 \n" +"2: cst.p %3,%M0 ,cc3,#1 \n" +" corcc gr29,gr29,gr0 ,cc3,#1 \n" /* clear ICC3.Z if store happens */ +" beq icc3,#0,0b \n" +" setlos 0,%2 \n" +"3: \n" +".subsection 2 \n" +"4: setlos %5,%2 \n" +" bra 3b \n" +".previous \n" +".section __ex_table,\"a\" \n" +" .balign 8 \n" +" .long 1b,4b \n" +" .long 2b,4b \n" +".previous" + :"+U" (*uaddr),"=&r" (oldval),"=&r" (ret),"=r" (oparg) + :"3" (oparg),"i" (-EFAULT ) + :"memory" ,"cc7" ,"cc3" ,"icc3" + ); + + *_oldval = oldval; +return ret; +} + +static inline int atomic_futex_op_xchg_xor (int oparg,int __user *uaddr,int *_oldval) +{ +int oldval, ret; + + asm("0: \n" +" orcc gr0,gr0,gr0,icc3 \n" /* set ICC3.Z */ +" ckeq icc3,cc7 \n" +"1: ld.p %M0,%1 \n" /* LD.P/ORCR must be atomic */ +" orcr cc7,cc7,cc3 \n" /* set CC3 to true */ +" xor %1,%3,%3 \n" +"2: cst.p %3,%M0 ,cc3,#1 \n" +" corcc gr29,gr29,gr0 ,cc3,#1 \n" /* clear ICC3.Z if store happens */ +" beq icc3,#0,0b \n" +" setlos 0,%2 \n" +"3: \n" +".subsection 2 \n" +"4: setlos %5,%2 \n" +" bra 3b \n" +".previous \n" +".section __ex_table,\"a\" \n" +" .balign 8 \n" +" .long 1b,4b \n" +" .long 2b,4b \n" +".previous" + :"+U" (*uaddr),"=&r" (oldval),"=&r" (ret),"=r" (oparg) + :"3" (oparg),"i" (-EFAULT ) + :"memory" ,"cc7" ,"cc3" ,"icc3" + ); + + *_oldval = oldval; +return ret; +} + +/*****************************************************************************/ +/* +* do the futex operations +*/ +int futex_atomic_op_inuser (int encoded_op,int __user *uaddr) +{ +int op = (encoded_op >>28 ) &7 ; +int cmp = (encoded_op >>24 ) &15 ; +int oparg = (encoded_op <<8 ) >>20 ; +int cmparg = (encoded_op <<20 ) >>20 ; +int oldval =0 , ret; + +if (encoded_op & (FUTEX_OP_OPARG_SHIFT <<28 )) + oparg =1 << oparg; + +if (!access_ok (VERIFY_WRITE , uaddr,sizeof (int ))) +return -EFAULT ; + +inc_preempt_count (); + +switch (op) { +case FUTEX_OP_SET : + ret =atomic_futex_op_xchg_set (oparg, uaddr, &oldval); +break ; +case FUTEX_OP_ADD : + ret =atomic_futex_op_xchg_add (oparg, uaddr, &oldval); +break ; +case FUTEX_OP_OR : + ret =atomic_futex_op_xchg_or (oparg, uaddr, &oldval); +break ; +case FUTEX_OP_ANDN : + ret =atomic_futex_op_xchg_and (~oparg, uaddr, &oldval); +break ; +case FUTEX_OP_XOR : + ret =atomic_futex_op_xchg_xor (oparg, uaddr, &oldval); +break ; +default : + ret = -ENOSYS ; +break ; + } + +dec_preempt_count (); + +if (!ret) { +switch (cmp) { +case FUTEX_OP_CMP_EQ : ret = (oldval == cmparg);break ; +case FUTEX_OP_CMP_NE : ret = (oldval != cmparg);break ; +case FUTEX_OP_CMP_LT : ret = (oldval < cmparg);break ; +case FUTEX_OP_CMP_GE : ret = (oldval >= cmparg);break ; +case FUTEX_OP_CMP_LE : ret = (oldval <= cmparg);break ; +case FUTEX_OP_CMP_GT : ret = (oldval > cmparg);break ; +default : ret = -ENOSYS ;break ; + } + } + +return ret; + +}/* end futex_atomic_op_inuser() */
arch/frv/kernel/Makefile +1 -0
@@ -20,3 +20,4 @@ obj-$(CONFIG_FUJITSU_MB93493) += irq-mb9 obj-$(CONFIG_PM) += pm.o cmode.o obj-$(CONFIG_MB93093_PDK) += pm-mb93093.o obj-$(CONFIG_SYSCTL) += sysctl.o +obj-$(CONFIG_FUTEX) += futex.o
include/asm-frv/futex.h +1 -41
48 lines not on this page · raw message
Quoted 3 lines · David Howells
(3) Not permitting a file to be shrunk if it would truncate any shared mappings (private mappings are copied).
Why?
Truncate is _supposed_ to get rid of any shared mmap stuff.
Linus
Quoted 6 lines · Linus Torvalds
Quoted 2 lines
(3) Not permitting a file to be shrunk if it would truncate any shared mappings (private mappings are copied).
Why?
Truncate is _supposed_ to get rid of any shared mmap stuff.
Yeah... but under _NOMMU_ conditions, it can't. There's no MMU around to enforce the fact that the mapping has been shrunk.
Imagine two processes: one creates a shmem file and makes it a certain size; both processes mmap it; then one of the processes attempts to shrink it. The process that shrank it knows it has been shortened and that the memory has been release, but the second process may not. Splat.
David
Quoted 4 lines · David Howells
Quoted 1 line
Truncate is _supposed_ to get rid of any shared mmap stuff.
Yeah... but under _NOMMU_ conditions, it can't. There's no MMU around to enforce the fact that the mapping has been shrunk.
I may not have mentioned, but this does not apply under MMU conditions. A different file in fs/ramfs/ swings into action and does the normal thing.
David