Permalink
Cannot retrieve contributors at this time
Name already in use
A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
pvfs2-osd/patches/posix-ext/posix-readwritex.patch
Go to fileThis commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
898 lines (873 sloc)
28 KB
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/arch/i386/kernel/syscall_table.S vanilla-new/arch/i386/kernel/syscall_table.S | |
| --- vanilla/arch/i386/kernel/syscall_table.S 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/arch/i386/kernel/syscall_table.S 2006-05-24 16:31:20.000000000 -0500 | |
| @@ -320,3 +320,5 @@ | |
| .long sys_fstatlite64 | |
| .long sys_getdents_plus | |
| .long sys_getdents64_plus /* 320 */ | |
| + .long sys_readx | |
| + .long sys_writex | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/arch/ia64/kernel/entry.S vanilla-new/arch/ia64/kernel/entry.S | |
| --- vanilla/arch/ia64/kernel/entry.S 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/arch/ia64/kernel/entry.S 2006-05-24 16:32:28.000000000 -0500 | |
| @@ -1626,5 +1626,7 @@ | |
| data8 sys_newfstatlite | |
| data8 sys_getdents_plus | |
| data8 sys_getdents64_plus | |
| + data8 sys_readx | |
| + data8 sys_writex | |
| .org sys_call_table + 8*NR_syscalls // guard against failures to increase NR_syscalls | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/arch/powerpc/kernel/systbl.S vanilla-new/arch/powerpc/kernel/systbl.S | |
| --- vanilla/arch/powerpc/kernel/systbl.S 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/arch/powerpc/kernel/systbl.S 2006-05-24 16:34:44.000000000 -0500 | |
| @@ -329,3 +329,5 @@ | |
| SYSCALL(newfstatlite) | |
| SYSCALL(sys_getdents_plus) | |
| SYSCALL(sys_getdents64_plus) | |
| +SYSCALL(sys_readx) | |
| +SYSCALL(sys_writex) | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/arch/x86_64/ia32/ia32entry.S vanilla-new/arch/x86_64/ia32/ia32entry.S | |
| --- vanilla/arch/x86_64/ia32/ia32entry.S 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/arch/x86_64/ia32/ia32entry.S 2006-05-24 16:32:01.000000000 -0500 | |
| @@ -698,6 +698,8 @@ | |
| .quad sys32_fstatlite64 | |
| .quad sys_ni_syscall /* getdents_plus */ | |
| .quad sys_ni_syscall /* getdents64_plus */ | |
| + .quad compat_sys_readx | |
| + .quad compat_sys_writex | |
| ia32_syscall_end: | |
| .rept IA32_NR_syscalls-(ia32_syscall_end-ia32_sys_call_table)/8 | |
| .quad ni_syscall | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/fs/compat.c vanilla-new/fs/compat.c | |
| --- vanilla/fs/compat.c 2006-05-09 18:53:48.000000000 -0500 | |
| +++ vanilla-new/fs/compat.c 2006-05-24 16:26:00.000000000 -0500 | |
| @@ -1344,6 +1344,322 @@ | |
| return ret; | |
| } | |
| +static ssize_t compat_do_readx_writex(int type, struct file *file, | |
| + const struct compat_iovec __user * uvector, | |
| + unsigned long nr_segs, | |
| + const struct compat_xtvec __user * xtuvector, | |
| + unsigned long xtnr_segs) | |
| +{ | |
| + typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *); | |
| + typedef ssize_t (*iov_fn_t)(struct file *, const struct iovec *, unsigned long, loff_t *); | |
| + typedef ssize_t (*iox_fn_t)(struct file *, const struct iovec *, unsigned long, | |
| + const struct xtvec *, unsigned long); | |
| + | |
| + compat_ssize_t tot_len, tot_xtlen; | |
| + struct iovec iovstack[UIO_FASTIOV]; | |
| + struct iovec *iov=iovstack; | |
| + struct xtvec xtvstack[UIO_FASTIOV]; | |
| + struct xtvec *xtv=xtvstack; | |
| + ssize_t ret; | |
| + int seg; | |
| + io_fn_t fn = NULL; | |
| + iov_fn_t fnv = NULL; | |
| + iox_fn_t fnx = NULL; | |
| + | |
| + /* | |
| + * readx does not make much sense if nr_segs <= 0 (OR) xtnr_segs <= 0 | |
| + * We return 0 similar to how readv/writev do. | |
| + */ | |
| + ret = 0; | |
| + if (nr_segs == 0 || xtnr_segs == 0) | |
| + goto out; | |
| + | |
| + /* | |
| + * First get the "struct iovec" from user memory and | |
| + * verify all the pointers | |
| + */ | |
| + ret = -EINVAL; | |
| + if ((nr_segs > UIO_MAXIOV) || (nr_segs <= 0)) | |
| + goto out; | |
| + if ((xtnr_segs > UIO_MAXIOV) || (xtnr_segs <= 0)) | |
| + goto out; | |
| + if (!file->f_op) | |
| + goto out; | |
| + if (nr_segs > UIO_FASTIOV) { | |
| + ret = -ENOMEM; | |
| + iov = kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!iov) | |
| + goto out; | |
| + } | |
| + if (xtnr_segs > UIO_FASTIOV) { | |
| + ret = -ENOMEM; | |
| + xtv = kmalloc(xtnr_segs * sizeof(struct xtvec), GFP_KERNEL); | |
| + if (!xtv) { | |
| + goto out; | |
| + } | |
| + } | |
| + ret = -EFAULT; | |
| + if (!access_ok(VERIFY_READ, uvector, nr_segs * sizeof(*uvector))) | |
| + goto out; | |
| + if (!access_ok(VERIFY_READ, xtuvector, xtnr_segs * sizeof(*xtuvector))) | |
| + goto out; | |
| + | |
| + /* | |
| + * Single unix specification: | |
| + * We should -EINVAL if an element length is not >= 0 and fitting an | |
| + * ssize_t. The total length is fitting an ssize_t | |
| + * | |
| + * Be careful here because iov_len is a size_t not an ssize_t | |
| + */ | |
| + tot_len = 0; | |
| + ret = -EINVAL; | |
| + for (seg = 0; seg < nr_segs; seg++) { | |
| + compat_ssize_t tmp = tot_len, len; | |
| + compat_uptr_t buf; | |
| + | |
| + if (__get_user(len, &uvector->iov_len) || | |
| + __get_user(buf, &uvector->iov_base)) { | |
| + ret = -EFAULT; | |
| + goto out; | |
| + } | |
| + if (len < 0) /* size_t not fitting an compat_ssize_t .. */ | |
| + goto out; | |
| + tot_len += len; | |
| + if (tot_len < tmp) /* maths overflow on the compat_ssize_t */ | |
| + goto out; | |
| + iov[seg].iov_base = compat_ptr(buf); | |
| + iov[seg].iov_len = (compat_size_t) len; | |
| + uvector++; | |
| + } | |
| + if (tot_len == 0) { | |
| + ret = 0; | |
| + goto out; | |
| + } | |
| + tot_xtlen = 0; | |
| + ret = -EINVAL; | |
| + for (seg = 0; seg < xtnr_segs; seg++) { | |
| + compat_ssize_t tmp = tot_xtlen, len; | |
| + compat_off_t off; | |
| + loff_t foff; | |
| + | |
| + if (__get_user(off, &xtuvector->xtv_off) || | |
| + __get_user(len, &xtuvector->xtv_len)) { | |
| + ret = -EFAULT; | |
| + goto out; | |
| + } | |
| + if (len < 0) /* size_t not fitting an compat_ssize_t .. */ | |
| + goto out; | |
| + if (off < 0) /* off_t not fitting an loff_t */ | |
| + goto out; | |
| + tot_xtlen += len; | |
| + if (tot_xtlen < tmp) /* overflow on the compat_ssize_t */ | |
| + goto out; | |
| + foff = (loff_t) off; | |
| + ret = rw_verify_area(type, file, &foff, len); | |
| + if (ret < 0) | |
| + goto out; | |
| + xtv[seg].xtv_off = (compat_off_t) off; | |
| + xtv[seg].xtv_len = (compat_size_t) len; | |
| + xtuvector++; | |
| + } | |
| + /* if sizes of file and mem don't match up, error out */ | |
| + if (tot_xtlen != tot_len) { | |
| + ret = -EINVAL; | |
| + goto out; | |
| + } | |
| + | |
| + if (type == READ) { | |
| + fn = (io_fn_t)file->f_op->read; | |
| + fnv = (iov_fn_t)file->f_op->readv; | |
| + fnx = (iox_fn_t)file->f_op->readx; | |
| + } else { | |
| + fn = (io_fn_t)file->f_op->write; | |
| + fnv = (iov_fn_t)file->f_op->writev; | |
| + fnx = (iox_fn_t)file->f_op->writex; | |
| + } | |
| + /* if we had a scatter-gather callback in memory and file, go for it */ | |
| + if (fnx) { | |
| + ret = fnx(file, iov, nr_segs, xtv, xtnr_segs); | |
| + goto out; | |
| + } | |
| + /* else try to do it by hand using readv/writev operations */ | |
| + else if (fnv) { | |
| + unsigned long xtiov_index = 0, op_iov_index = 0, iov_index = 0; | |
| + struct iovec *op_iov = NULL, *copied_iovector = NULL; | |
| + struct xtvec *copied_xtvector = NULL; | |
| + | |
| + ret = -ENOMEM; | |
| + op_iov = (struct iovec *) kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!op_iov) | |
| + goto err_out1; | |
| + copied_iovector = (struct iovec *) kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!copied_iovector) | |
| + goto err_out1; | |
| + copied_xtvector = (struct xtvec *) kmalloc(xtnr_segs * sizeof(struct xtvec), GFP_KERNEL); | |
| + if (!copied_xtvector) | |
| + goto err_out1; | |
| + memcpy(copied_iovector, iov, nr_segs * sizeof(struct iovec)); | |
| + memcpy(copied_xtvector, xtv, xtnr_segs * sizeof(struct xtvec)); | |
| + ret = 0; | |
| + iov_index = 0; | |
| + for (xtiov_index = 0; xtiov_index < xtnr_segs; xtiov_index++) { | |
| + loff_t pos; | |
| + ssize_t nr, tot_nr; | |
| + | |
| + pos = copied_xtvector[xtiov_index].xtv_off; | |
| + op_iov_index = 0; | |
| + tot_nr = 0; | |
| + | |
| + /* Finish an entire stream and .. */ | |
| + while (copied_xtvector[xtiov_index].xtv_len > 0) { | |
| + size_t min_len; | |
| + if (unlikely((iov_index >= nr_segs) || (op_iov_index >= nr_segs))) { | |
| + printk(KERN_ERR "iov_index %ld or op_iov_index %ld cannot exceed number of iov segments (%ld)\n", | |
| + iov_index, op_iov_index, nr_segs); | |
| + ret = -EINVAL; | |
| + goto err_out1; | |
| + } | |
| + min_len = min(copied_xtvector[xtiov_index].xtv_len, copied_iovector[iov_index].iov_len); | |
| + op_iov[op_iov_index].iov_base = copied_iovector[iov_index].iov_base; | |
| + op_iov[op_iov_index++].iov_len = min_len; | |
| + copied_xtvector[xtiov_index].xtv_len -= min_len; | |
| + copied_iovector[iov_index].iov_len -= min_len; | |
| + copied_iovector[iov_index].iov_base += min_len; | |
| + tot_nr += min_len; | |
| + /* Advance memory stream if we have exhausted it */ | |
| + if (copied_iovector[iov_index].iov_len <= 0) { | |
| + iov_index++; | |
| + } | |
| + } | |
| + /* .. issue a vectored operation for that region */ | |
| + nr = fnv(file, op_iov, op_iov_index, &pos); | |
| + if (nr < 0) { | |
| + if (!ret) ret = nr; | |
| + break; | |
| + } | |
| + ret += nr; | |
| + if (nr != tot_nr) | |
| + break; | |
| + } | |
| +err_out1: | |
| + kfree(op_iov); | |
| + kfree(copied_iovector); | |
| + kfree(copied_xtvector); | |
| + goto out; | |
| + } | |
| + /* Do it by hand, with plain read/write operations */ | |
| + else { | |
| + unsigned long mem_ct = 0, str_ct = 0; | |
| + struct xtvec *copied_xtvector = NULL; | |
| + struct iovec *copied_iovector = NULL; | |
| + | |
| + ret = -ENOMEM; | |
| + copied_iovector = (struct iovec *) kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!copied_iovector) | |
| + goto err_out2; | |
| + copied_xtvector = (struct xtvec *) kmalloc(xtnr_segs * sizeof(struct xtvec), GFP_KERNEL); | |
| + if (!copied_xtvector) | |
| + goto err_out2; | |
| + memcpy(copied_iovector, iov, nr_segs * sizeof(struct iovec)); | |
| + memcpy(copied_xtvector, xtv, xtnr_segs * sizeof(struct xtvec)); | |
| + | |
| + ret = 0; | |
| + mem_ct = 0; | |
| + str_ct = 0; | |
| + while ((mem_ct < nr_segs) && (str_ct < xtnr_segs)) { | |
| + size_t min_len; | |
| + loff_t pos; | |
| + ssize_t nr; | |
| + void __user *base; | |
| + | |
| + pos = copied_xtvector[str_ct].xtv_off; | |
| + base = copied_iovector[mem_ct].iov_base; | |
| + min_len = min(copied_xtvector[str_ct].xtv_len, copied_iovector[mem_ct].iov_len); | |
| + copied_xtvector[str_ct].xtv_len -= min_len; | |
| + copied_xtvector[str_ct].xtv_off += min_len; | |
| + copied_iovector[mem_ct].iov_len -= min_len; | |
| + copied_iovector[mem_ct].iov_base += min_len; | |
| + if (copied_iovector[mem_ct].iov_len <= 0) | |
| + mem_ct++; | |
| + if (copied_xtvector[str_ct].xtv_len <= 0) | |
| + str_ct++; | |
| + /* Issue the smallest region that is contiguous in memory and on file */ | |
| + nr = fn(file, base, min_len, &pos); | |
| + if (nr < 0) { | |
| + if (!ret) ret = nr; | |
| + break; | |
| + } | |
| + ret += nr; | |
| + if (nr != min_len) | |
| + break; | |
| + } | |
| +err_out2: | |
| + kfree(copied_xtvector); | |
| + kfree(copied_iovector); | |
| + } | |
| +out: | |
| + if (iov != iovstack) | |
| + kfree(iov); | |
| + if (xtv != xtvstack) | |
| + kfree(xtv); | |
| + if ((ret + (type == READ)) > 0) { | |
| + if (type == READ) | |
| + fsnotify_access(file->f_dentry); | |
| + else | |
| + fsnotify_modify(file->f_dentry); | |
| + } | |
| + return ret; | |
| +} | |
| + | |
| +asmlinkage ssize_t | |
| +compat_sys_readx(unsigned long fd, const struct compat_iovec __user *vec, unsigned long vlen, | |
| + const struct compat_xtvec __user *xtv, unsigned long xtvlen) | |
| +{ | |
| + struct file *file; | |
| + ssize_t ret = -EBADF; | |
| + | |
| + file = fget(fd); | |
| + if (!file) | |
| + return -EBADF; | |
| + | |
| + if (!(file->f_mode & FMODE_READ)) | |
| + goto out; | |
| + | |
| + ret = -EINVAL; | |
| + if (!file->f_op || (!file->f_op->readx && !file->f_op->readv && !file->f_op->read)) | |
| + goto out; | |
| + | |
| + ret = compat_do_readx_writex(READ, file, vec, vlen, xtv, xtvlen); | |
| + | |
| +out: | |
| + fput(file); | |
| + return ret; | |
| +} | |
| + | |
| +asmlinkage ssize_t | |
| +compat_sys_writex(unsigned long fd, const struct compat_iovec __user *vec, unsigned long vlen, | |
| + const struct compat_xtvec __user *xtv, unsigned long xtvlen) | |
| +{ | |
| + struct file *file; | |
| + ssize_t ret = -EBADF; | |
| + | |
| + file = fget(fd); | |
| + if (!file) | |
| + return -EBADF; | |
| + if (!(file->f_mode & FMODE_WRITE)) | |
| + goto out; | |
| + | |
| + ret = -EINVAL; | |
| + if (!file->f_op || (!file->f_op->writex && !file->f_op->writev && !file->f_op->write)) | |
| + goto out; | |
| + | |
| + ret = compat_do_readx_writex(WRITE, file, vec, vlen, xtv, xtvlen); | |
| + | |
| +out: | |
| + fput(file); | |
| + return ret; | |
| +} | |
| + | |
| /* | |
| * Exactly like fs/open.c:sys_open(), except that it doesn't set the | |
| * O_LARGEFILE flag. | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/fs/read_write.c vanilla-new/fs/read_write.c | |
| --- vanilla/fs/read_write.c 2006-03-19 23:53:29.000000000 -0600 | |
| +++ vanilla-new/fs/read_write.c 2006-05-24 16:22:20.000000000 -0500 | |
| @@ -374,6 +374,8 @@ | |
| return ret; | |
| } | |
| +EXPORT_SYMBOL_GPL(sys_write); | |
| + | |
| asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf, | |
| size_t count, loff_t pos) | |
| { | |
| @@ -636,6 +638,334 @@ | |
| return ret; | |
| } | |
| +static ssize_t do_readx_writex(int type, struct file *file, | |
| + const struct iovec __user * uvector, | |
| + unsigned long nr_segs, | |
| + const struct xtvec __user * xtuvector, | |
| + unsigned long xtnr_segs) | |
| +{ | |
| + typedef ssize_t (*io_fn_t)(struct file *, char __user *, size_t, loff_t *); | |
| + typedef ssize_t (*iov_fn_t)(struct file *, const struct iovec *, unsigned long, loff_t *); | |
| + typedef ssize_t (*iox_fn_t)(struct file *, const struct iovec *, unsigned long, | |
| + const struct xtvec *, unsigned long); | |
| + | |
| + size_t tot_len, tot_xtlen; | |
| + struct iovec iovstack[UIO_FASTIOV]; | |
| + struct iovec *iov=iovstack; | |
| + struct xtvec xtvstack[UIO_FASTIOV]; | |
| + struct xtvec *xtv=xtvstack; | |
| + ssize_t ret; | |
| + int seg; | |
| + io_fn_t fn = NULL; | |
| + iov_fn_t fnv = NULL; | |
| + iox_fn_t fnx = NULL; | |
| + | |
| + /* | |
| + * readx does not make much sense if nr_segs <= 0 (OR) xtnr_segs <= 0 | |
| + * We return 0 similar to how readv/writev do. | |
| + */ | |
| + ret = 0; | |
| + if (nr_segs == 0 || xtnr_segs == 0) | |
| + goto out; | |
| + | |
| + /* | |
| + * First get the "struct iovec" from user memory and | |
| + * verify all the pointers | |
| + */ | |
| + ret = -EINVAL; | |
| + if ((nr_segs > UIO_MAXIOV) || (nr_segs <= 0)) | |
| + goto out; | |
| + if ((xtnr_segs > UIO_MAXIOV) || (xtnr_segs <= 0)) | |
| + goto out; | |
| + if (!file->f_op) | |
| + goto out; | |
| + if (nr_segs > UIO_FASTIOV) { | |
| + ret = -ENOMEM; | |
| + iov = kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!iov) | |
| + goto out; | |
| + } | |
| + if (xtnr_segs > UIO_FASTIOV) { | |
| + ret = -ENOMEM; | |
| + xtv = kmalloc(xtnr_segs * sizeof(struct xtvec), GFP_KERNEL); | |
| + if (!xtv) { | |
| + goto out; | |
| + } | |
| + } | |
| + ret = -EFAULT; | |
| + if (copy_from_user(iov, uvector, nr_segs * sizeof(*uvector))) | |
| + goto out; | |
| + if (copy_from_user(xtv, xtuvector, xtnr_segs * sizeof(*xtuvector))) | |
| + goto out; | |
| + | |
| + /* | |
| + * Single unix specification: | |
| + * We should -EINVAL if an element length is not >= 0 and fitting an | |
| + * ssize_t. The total length is fitting an ssize_t | |
| + * | |
| + * Be careful here because iov_len is a size_t not an ssize_t | |
| + */ | |
| + tot_len = 0; | |
| + ret = -EINVAL; | |
| + for (seg = 0; seg < nr_segs; seg++) { | |
| + void __user *buf = iov[seg].iov_base; | |
| + ssize_t len = (ssize_t)iov[seg].iov_len; | |
| + | |
| + if (len < 0) /* size_t not fitting an ssize_t .. */ | |
| + goto out; | |
| + if (unlikely(!access_ok(vrfy_dir(type), buf, len))) | |
| + goto Efault; | |
| + tot_len += len; | |
| + if ((ssize_t)tot_len < 0) /* maths overflow on the ssize_t */ | |
| + goto out; | |
| + } | |
| + if (tot_len == 0) { | |
| + ret = 0; | |
| + goto out; | |
| + } | |
| + tot_xtlen = 0; | |
| + ret = -EINVAL; | |
| + for (seg = 0; seg < xtnr_segs; seg++) { | |
| + loff_t off = (loff_t) xtv[seg].xtv_off; | |
| + ssize_t len = (ssize_t)xtv[seg].xtv_len; | |
| + | |
| + if (len < 0) /* size_t not fitting an ssize_t .. */ | |
| + goto out; | |
| + if (off < 0) /* off_t not fitting an loff_t */ | |
| + goto out; | |
| + tot_xtlen += len; | |
| + if ((ssize_t)tot_xtlen < 0) /* overflow on the ssize_t */ | |
| + goto out; | |
| + ret = rw_verify_area(type, file, &off, len); | |
| + if (ret < 0) | |
| + goto out; | |
| + } | |
| + /* if sizes of file and mem don't match up, error out */ | |
| + if (tot_xtlen != tot_len) { | |
| + ret = -EINVAL; | |
| + goto out; | |
| + } | |
| + | |
| + ret = security_file_permission(file, type == READ ? MAY_READ : MAY_WRITE); | |
| + if (ret) | |
| + goto out; | |
| + | |
| + if (type == READ) { | |
| + fn = (io_fn_t)file->f_op->read; | |
| + fnv = (iov_fn_t)file->f_op->readv; | |
| + fnx = (iox_fn_t)file->f_op->readx; | |
| + } else { | |
| + fn = (io_fn_t)file->f_op->write; | |
| + fnv = (iov_fn_t)file->f_op->writev; | |
| + fnx = (iox_fn_t)file->f_op->writex; | |
| + } | |
| + /* if we had a scatter-gather callback in memory and file, go for it */ | |
| + if (fnx) { | |
| + ret = fnx(file, iov, nr_segs, xtv, xtnr_segs); | |
| + goto out; | |
| + } | |
| + /* else try to do it by hand using readv/writev operations */ | |
| + else if (fnv) { | |
| + unsigned long xtiov_index = 0, op_iov_index = 0, iov_index = 0; | |
| + struct iovec *op_iov = NULL, *copied_iovector = NULL; | |
| + struct xtvec *copied_xtvector = NULL; | |
| + | |
| + ret = -ENOMEM; | |
| + op_iov = (struct iovec *) kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!op_iov) | |
| + goto err_out1; | |
| + copied_iovector = (struct iovec *) kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!copied_iovector) | |
| + goto err_out1; | |
| + copied_xtvector = (struct xtvec *) kmalloc(xtnr_segs * sizeof(struct xtvec), GFP_KERNEL); | |
| + if (!copied_xtvector) | |
| + goto err_out1; | |
| + memcpy(copied_iovector, iov, nr_segs * sizeof(struct iovec)); | |
| + memcpy(copied_xtvector, xtv, xtnr_segs * sizeof(struct xtvec)); | |
| + ret = 0; | |
| + iov_index = 0; | |
| + for (xtiov_index = 0; xtiov_index < xtnr_segs; xtiov_index++) { | |
| + loff_t pos; | |
| + ssize_t nr, tot_nr; | |
| + | |
| + pos = copied_xtvector[xtiov_index].xtv_off; | |
| + op_iov_index = 0; | |
| + tot_nr = 0; | |
| + | |
| + /* Finish an entire stream and .. */ | |
| + while (copied_xtvector[xtiov_index].xtv_len > 0) { | |
| + size_t min_len; | |
| + if (unlikely((iov_index >= nr_segs) || (op_iov_index >= nr_segs))) { | |
| + printk(KERN_ERR "iov_index %ld or op_iov_index %ld cannot exceed number of iov segments (%ld)\n", | |
| + iov_index, op_iov_index, nr_segs); | |
| + ret = -EINVAL; | |
| + goto err_out1; | |
| + } | |
| + min_len = min(copied_xtvector[xtiov_index].xtv_len, copied_iovector[iov_index].iov_len); | |
| + op_iov[op_iov_index].iov_base = copied_iovector[iov_index].iov_base; | |
| + op_iov[op_iov_index++].iov_len = min_len; | |
| + copied_xtvector[xtiov_index].xtv_len -= min_len; | |
| + copied_iovector[iov_index].iov_len -= min_len; | |
| + copied_iovector[iov_index].iov_base += min_len; | |
| + tot_nr += min_len; | |
| + /* Advance memory stream if we have exhausted it */ | |
| + if (copied_iovector[iov_index].iov_len <= 0) { | |
| + iov_index++; | |
| + } | |
| + } | |
| + /* .. issue a vectored operation for that region */ | |
| + nr = fnv(file, op_iov, op_iov_index, &pos); | |
| + if (nr < 0) { | |
| + if (!ret) ret = nr; | |
| + break; | |
| + } | |
| + ret += nr; | |
| + if (nr != tot_nr) | |
| + break; | |
| + } | |
| +err_out1: | |
| + kfree(op_iov); | |
| + kfree(copied_iovector); | |
| + kfree(copied_xtvector); | |
| + goto out; | |
| + } | |
| + /* Do it by hand, with plain read/write operations */ | |
| + else { | |
| + unsigned long mem_ct = 0, str_ct = 0; | |
| + struct xtvec *copied_xtvector = NULL; | |
| + struct iovec *copied_iovector = NULL; | |
| + | |
| + ret = -ENOMEM; | |
| + copied_iovector = (struct iovec *) kmalloc(nr_segs * sizeof(struct iovec), GFP_KERNEL); | |
| + if (!copied_iovector) | |
| + goto err_out2; | |
| + copied_xtvector = (struct xtvec *) kmalloc(xtnr_segs * sizeof(struct xtvec), GFP_KERNEL); | |
| + if (!copied_xtvector) | |
| + goto err_out2; | |
| + memcpy(copied_iovector, iov, nr_segs * sizeof(struct iovec)); | |
| + memcpy(copied_xtvector, xtv, xtnr_segs * sizeof(struct xtvec)); | |
| + | |
| + ret = 0; | |
| + mem_ct = 0; | |
| + str_ct = 0; | |
| + while ((mem_ct < nr_segs) && (str_ct < xtnr_segs)) { | |
| + size_t min_len; | |
| + loff_t pos; | |
| + ssize_t nr; | |
| + void __user *base; | |
| + | |
| + pos = copied_xtvector[str_ct].xtv_off; | |
| + base = copied_iovector[mem_ct].iov_base; | |
| + min_len = min(copied_xtvector[str_ct].xtv_len, copied_iovector[mem_ct].iov_len); | |
| + copied_xtvector[str_ct].xtv_len -= min_len; | |
| + copied_xtvector[str_ct].xtv_off += min_len; | |
| + copied_iovector[mem_ct].iov_len -= min_len; | |
| + copied_iovector[mem_ct].iov_base += min_len; | |
| + if (copied_iovector[mem_ct].iov_len <= 0) | |
| + mem_ct++; | |
| + if (copied_xtvector[str_ct].xtv_len <= 0) | |
| + str_ct++; | |
| + /* Issue the smallest region that is contiguous in memory and on file */ | |
| + nr = fn(file, base, min_len, &pos); | |
| + if (nr < 0) { | |
| + if (!ret) ret = nr; | |
| + break; | |
| + } | |
| + ret += nr; | |
| + if (nr != min_len) | |
| + break; | |
| + } | |
| +err_out2: | |
| + kfree(copied_xtvector); | |
| + kfree(copied_iovector); | |
| + } | |
| +out: | |
| + if (iov != iovstack) | |
| + kfree(iov); | |
| + if (xtv != xtvstack) | |
| + kfree(xtv); | |
| + if ((ret + (type == READ)) > 0) { | |
| + if (type == READ) | |
| + fsnotify_access(file->f_dentry); | |
| + else | |
| + fsnotify_modify(file->f_dentry); | |
| + } | |
| + return ret; | |
| +Efault: | |
| + ret = -EFAULT; | |
| + goto out; | |
| +} | |
| + | |
| +ssize_t vfs_readx(struct file *file, const struct iovec __user *vec, | |
| + unsigned long vlen, const struct xtvec __user *xtvec, unsigned long xtvlen) | |
| +{ | |
| + if (!(file->f_mode & FMODE_READ)) | |
| + return -EBADF; | |
| + if (!file->f_op || (!file->f_op->readx && !file->f_op->readv && !file->f_op->read)) | |
| + return -EINVAL; | |
| + | |
| + return do_readx_writex(READ, file, vec, vlen, xtvec, xtvlen); | |
| +} | |
| + | |
| +EXPORT_SYMBOL_GPL(vfs_readx); | |
| + | |
| +ssize_t vfs_writex(struct file *file, const struct iovec __user *vec, | |
| + unsigned long vlen, const struct xtvec __user *xtvec, unsigned long xtvlen) | |
| +{ | |
| + if (!(file->f_mode & FMODE_WRITE)) | |
| + return -EBADF; | |
| + if (!file->f_op || (!file->f_op->writex && !file->f_op->writev && !file->f_op->write)) | |
| + return -EINVAL; | |
| + | |
| + return do_readx_writex(WRITE, file, vec, vlen, xtvec, xtvlen); | |
| +} | |
| + | |
| +EXPORT_SYMBOL_GPL(vfs_writex); | |
| + | |
| +asmlinkage ssize_t sys_readx(unsigned long fd, | |
| + const struct iovec __user *vec, | |
| + unsigned long vlen, | |
| + const struct xtvec __user *xtvec, | |
| + unsigned long xtvlen) | |
| +{ | |
| + struct file *file; | |
| + ssize_t ret = -EBADF; | |
| + int fput_needed; | |
| + | |
| + file = fget_light(fd, &fput_needed); | |
| + if (file) { | |
| + ret = vfs_readx(file, vec, vlen, xtvec, xtvlen); | |
| + fput_light(file, fput_needed); | |
| + } | |
| + | |
| + if (ret > 0) | |
| + current->rchar += ret; | |
| + current->syscr++; | |
| + return ret; | |
| +} | |
| + | |
| +asmlinkage ssize_t sys_writex(unsigned long fd, | |
| + const struct iovec __user *vec, | |
| + unsigned long vlen, | |
| + const struct xtvec __user *xtvec, | |
| + unsigned long xtvlen) | |
| +{ | |
| + struct file *file; | |
| + ssize_t ret = -EBADF; | |
| + int fput_needed; | |
| + | |
| + file = fget_light(fd, &fput_needed); | |
| + if (file) { | |
| + ret = vfs_writex(file, vec, vlen, xtvec, xtvlen); | |
| + fput_light(file, fput_needed); | |
| + } | |
| + | |
| + if (ret > 0) | |
| + current->wchar += ret; | |
| + current->syscw++; | |
| + return ret; | |
| +} | |
| + | |
| static ssize_t do_sendfile(int out_fd, int in_fd, loff_t *ppos, | |
| size_t count, loff_t max) | |
| { | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/asm-i386/unistd.h vanilla-new/include/asm-i386/unistd.h | |
| --- vanilla/include/asm-i386/unistd.h 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/include/asm-i386/unistd.h 2006-05-24 16:27:28.000000000 -0500 | |
| @@ -326,8 +326,10 @@ | |
| #define __NR_fstatlite64 318 | |
| #define __NR_getdents_plus 319 | |
| #define __NR_getdents64_plus 320 | |
| +#define __NR_readx 321 | |
| +#define __NR_writex 322 | |
| -#define NR_syscalls 321 | |
| +#define NR_syscalls 323 | |
| /* | |
| * user-visible error numbers are in the range -1 - -128: see | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/asm-ia64/unistd.h vanilla-new/include/asm-ia64/unistd.h | |
| --- vanilla/include/asm-ia64/unistd.h 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/include/asm-ia64/unistd.h 2006-05-24 16:29:19.000000000 -0500 | |
| @@ -292,12 +292,14 @@ | |
| #define __NR_fstatlite 1301 | |
| #define __NR_getdents_plus 1302 | |
| #define __NR_getdents64_plus 1303 | |
| +#define __NR_readx 1304 | |
| +#define __NR_writex 1305 | |
| #ifdef __KERNEL__ | |
| #include <linux/config.h> | |
| -#define NR_syscalls 280 /* length of syscall table */ | |
| +#define NR_syscalls 282 /* length of syscall table */ | |
| #define __ARCH_WANT_SYS_RT_SIGACTION | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/asm-powerpc/unistd.h vanilla-new/include/asm-powerpc/unistd.h | |
| --- vanilla/include/asm-powerpc/unistd.h 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/include/asm-powerpc/unistd.h 2006-05-24 16:30:04.000000000 -0500 | |
| @@ -308,8 +308,10 @@ | |
| #define __NR_fstatlite 287 | |
| #define __NR_getdents_plus 288 | |
| #define __NR_getdents64_plus 289 | |
| +#define __NR_readx 290 | |
| +#define __NR_writex 291 | |
| -#define __NR_syscalls 290 | |
| +#define __NR_syscalls 292 | |
| #ifdef __KERNEL__ | |
| #define __NR__exit __NR_exit | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/asm-x86_64/unistd.h vanilla-new/include/asm-x86_64/unistd.h | |
| --- vanilla/include/asm-x86_64/unistd.h 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/include/asm-x86_64/unistd.h 2006-05-24 16:28:26.000000000 -0500 | |
| @@ -619,8 +619,12 @@ | |
| __SYSCALL(__NR_getdents_plus, sys_getdents_plus) | |
| #define __NR_getdents64_plus 279 | |
| __SYSCALL(__NR_getdents64_plus, sys_getdents64_plus) | |
| +#define __NR_readx 280 | |
| +__SYSCALL(__NR_readx, sys_readx) | |
| +#define __NR_writex 281 | |
| +__SYSCALL(__NR_writex, sys_writex) | |
| -#define __NR_syscall_max __NR_getdents64_plus | |
| +#define __NR_syscall_max __NR_writex | |
| #ifndef __NO_STUBS | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/linux/compat.h vanilla-new/include/linux/compat.h | |
| --- vanilla/include/linux/compat.h 2006-05-09 18:53:48.000000000 -0500 | |
| +++ vanilla-new/include/linux/compat.h 2006-05-24 16:26:29.000000000 -0500 | |
| @@ -61,6 +61,11 @@ | |
| compat_size_t iov_len; | |
| }; | |
| +struct compat_xtvec { | |
| + compat_off_t xtv_off; | |
| + compat_size_t xtv_len; | |
| +}; | |
| + | |
| struct compat_rlimit { | |
| compat_ulong_t rlim_cur; | |
| compat_ulong_t rlim_max; | |
| @@ -141,6 +146,13 @@ | |
| asmlinkage ssize_t compat_sys_writev(unsigned long fd, | |
| const struct compat_iovec __user *vec, unsigned long vlen); | |
| +asmlinkage ssize_t compat_sys_readx(unsigned long fd, | |
| + const struct compat_iovec __user *vec, unsigned long vlen, | |
| + const struct compat_xtvec __user *xtvec, unsigned long xtvlen); | |
| +asmlinkage ssize_t compat_sys_writex(unsigned long fd, | |
| + const struct compat_iovec __user *vec, unsigned long vlen, | |
| + const struct compat_xtvec __user *xtvec, unsigned long xtvlen); | |
| + | |
| int compat_do_execve(char * filename, compat_uptr_t __user *argv, | |
| compat_uptr_t __user *envp, struct pt_regs * regs); | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/linux/fs.h vanilla-new/include/linux/fs.h | |
| --- vanilla/include/linux/fs.h 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/include/linux/fs.h 2006-05-24 16:20:15.000000000 -0500 | |
| @@ -226,6 +226,7 @@ | |
| struct hd_geometry; | |
| struct iovec; | |
| +struct xtvec; | |
| struct nameidata; | |
| struct kiocb; | |
| struct pipe_inode_info; | |
| @@ -1048,6 +1049,8 @@ | |
| int (*lock) (struct file *, int, struct file_lock *); | |
| ssize_t (*readv) (struct file *, const struct iovec *, unsigned long, loff_t *); | |
| ssize_t (*writev) (struct file *, const struct iovec *, unsigned long, loff_t *); | |
| + ssize_t (*readx) (struct file *, const struct iovec *, unsigned long, const struct xtvec *, unsigned long); | |
| + ssize_t (*writex) (struct file *, const struct iovec *, unsigned long, const struct xtvec *, unsigned long); | |
| ssize_t (*sendfile) (struct file *, loff_t *, size_t, read_actor_t, void *); | |
| ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int); | |
| unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
| @@ -1092,6 +1095,11 @@ | |
| unsigned long, loff_t *); | |
| extern ssize_t vfs_writev(struct file *, const struct iovec __user *, | |
| unsigned long, loff_t *); | |
| +extern ssize_t vfs_readx(struct file *, const struct iovec __user *, | |
| + unsigned long, const struct xtvec __user *, unsigned long); | |
| +extern ssize_t vfs_writex(struct file *, const struct iovec __user *, | |
| + unsigned long, const struct xtvec __user *, unsigned long); | |
| + | |
| /* | |
| * NOTE: write_inode, delete_inode, clear_inode, put_inode can be called | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/linux/syscalls.h vanilla-new/include/linux/syscalls.h | |
| --- vanilla/include/linux/syscalls.h 2006-05-24 16:19:07.000000000 -0500 | |
| +++ vanilla-new/include/linux/syscalls.h 2006-05-24 16:23:31.000000000 -0500 | |
| @@ -17,6 +17,7 @@ | |
| struct iocb; | |
| struct io_event; | |
| struct iovec; | |
| +struct xtvec; | |
| struct itimerspec; | |
| struct itimerval; | |
| struct kexec_segment; | |
| @@ -402,6 +403,16 @@ | |
| asmlinkage ssize_t sys_writev(unsigned long fd, | |
| const struct iovec __user *vec, | |
| unsigned long vlen); | |
| +asmlinkage ssize_t sys_readx(unsigned long fd, | |
| + const struct iovec __user *vec, | |
| + unsigned long vlen, | |
| + const struct xtvec __user *xtvec, | |
| + unsigned long xtvlen); | |
| +asmlinkage ssize_t sys_writex(unsigned long fd, | |
| + const struct iovec __user *vec, | |
| + unsigned long vlen, | |
| + const struct xtvec __user *xtvec, | |
| + unsigned long xtvlen); | |
| asmlinkage ssize_t sys_pread64(unsigned int fd, char __user *buf, | |
| size_t count, loff_t pos); | |
| asmlinkage ssize_t sys_pwrite64(unsigned int fd, const char __user *buf, | |
| diff -Naur --exclude-from=/home/vilayann/redhat/BUILD/kernel-2.6.16/exclude vanilla/include/linux/uio.h vanilla-new/include/linux/uio.h | |
| --- vanilla/include/linux/uio.h 2006-03-19 23:53:29.000000000 -0600 | |
| +++ vanilla-new/include/linux/uio.h 2006-05-24 16:24:20.000000000 -0500 | |
| @@ -23,6 +23,12 @@ | |
| __kernel_size_t iov_len; /* Must be size_t (1003.1g) */ | |
| }; | |
| +struct xtvec | |
| +{ | |
| + __kernel_off_t xtv_off; /* must be off_t */ | |
| + __kernel_size_t xtv_len; /* must be size_t */ | |
| +}; | |
| + | |
| #ifdef __KERNEL__ | |
| struct kvec { | |
| @@ -30,6 +36,12 @@ | |
| size_t iov_len; | |
| }; | |
| +struct kxtvec | |
| +{ | |
| + off_t xtv_off; | |
| + size_t xtv_len; | |
| +}; | |
| + | |
| #endif | |
| /* |