In working on thread core dumps I've stumbled across a minor bug in the
generic `fork' code. The problem code is in routine `copy_mm' in
`kernel/fork.c':
/* Copy the current MM stuff.. */
memcpy(mm, oldmm, sizeof(*mm));
.
.
.
if (retval)
goto free_pt;
/*
* child gets a private LDT (if there was an LDT in the parent)
*/
copy_segments(tsk, mm);
.
.
.
free_pt:
mmput(mm);
The new `mm' doesn't get it's own private version of it's `context'
field until after the call to `copy_segments'. At the `goto' the pointer
`mm' points to a copy of the old `mm_struct', including a copy of the
`context' field. `mmput' will call `release_segments' which will free
the memory pointed to by the `context' field. Later, when `oldmm' is
freed, the kernel will try to free the same memory twice - very bad.
Admittedly, this will only occur on an obscure error condition but
it should be fixed. The attached patch fixes the problem by zeroing
out the `context' field immediately after the `mm' structure is
copied.
-- Don Dugger "Censeo Toto nos in Kansa esse decisse." - D. Gale n0ano@valinux.com Ph: 303/938-9838--- linux-2.4.4-ref/kernel/fork.c Thu Apr 26 07:11:17 2001 +++ linux/kernel/fork.c Wed May 2 14:39:38 2001 @@ -311,6 +311,10 @@ /* Copy the current MM stuff.. */ memcpy(mm, oldmm, sizeof(*mm)); + + /* Clear new context for now */ + memset(&mm->context, 0, sizeof(mm->context)); + if (!mm_init(mm)) goto fail_nomem; - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon May 07 2001 - 21:00:14 EST