Unfortunately the patch I submitted for shmdt has been overlooked.
It's not a very exciting patch, but it does bring the code into agreement
with the manpage. Here is an version of the patch updated for 2.4.16.
--- linux-2.4.16old/ipc/shm.c Mon Dec 10 11:53:12 2001
+++ linux-2.4.16/ipc/shm.c Mon Dec 10 15:08:51 2001
@@ -651,16 +651,19 @@
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *shmd, *shmdnext;
+ int retcode=-EINVAL;
down_write(&mm->mmap_sem);
for (shmd = mm->mmap; shmd; shmd = shmdnext) {
shmdnext = shmd->vm_next;
if (shmd->vm_ops == &shm_vm_ops
- && shmd->vm_start - (shmd->vm_pgoff << PAGE_SHIFT) ==
(ulong) shmaddr)
+ && shmd->vm_start - (shmd->vm_pgoff << PAGE_SHIFT) ==
(ulong) shmaddr) {
do_munmap(mm, shmd->vm_start, shmd->vm_end -
shmd->vm_start);
+ retcode=0;
+ }
}
up_write(&mm->mmap_sem);
- return 0;
+ return retcode;
}
#ifdef CONFIG_PROC_FS
From: <[email protected]>
Unfortunately the patch I submitted for shmdt has been overlooked.
It's not a very exciting patch, but it does bring the code into agreement
with the manpage. Here is an version of the patch updated for 2.4.16.
Yes, I agree. Referring to POSIX would have been more convincing,
but it is true that shmdt is supposed to return EINVAL when the
memory to be detached is not found.
Andries
--- linux-2.4.16old/ipc/shm.c Mon Dec 10 11:53:12 2001
+++ linux-2.4.16/ipc/shm.c Mon Dec 10 15:08:51 2001
@@ -651,16 +651,19 @@
{
struct mm_struct *mm = current->mm;
struct vm_area_struct *shmd, *shmdnext;
+ int retcode = -EINVAL;
down_write(&mm->mmap_sem);
for (shmd = mm->mmap; shmd; shmd = shmdnext) {
shmdnext = shmd->vm_next;
if (shmd->vm_ops == &shm_vm_ops
- && shmd->vm_start - (shmd->vm_pgoff << PAGE_SHIFT) == (ulong) shmaddr)
+ && shmd->vm_start - (shmd->vm_pgoff << PAGE_SHIFT) == (ulong) shmaddr) {
do_munmap(mm, shmd->vm_start, shmd->vm_end - shmd->vm_start);
+ retcode = 0;
+ }
}
up_write(&mm->mmap_sem);
- return 0;
+ return retcode;
}
#ifdef CONFIG_PROC_FS