![Guillaume Autran](/assets/img/avatar_default.png)
Fix for a race condition when a task gets preempted by another task while executing the destroy_context(...) in a FEW_CONTEXTS environment. mm->context == NO_CONTEXT but the context_map may indicate all contexts are in use. The solution to this problem is to disable kernel preemption while destroying a MMU context. Signed-off-by: Guillaume Autran <gautran@mrv.com> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
200 lines
5.6 KiB
C
200 lines
5.6 KiB
C
#ifdef __KERNEL__
|
|
#ifndef __PPC_MMU_CONTEXT_H
|
|
#define __PPC_MMU_CONTEXT_H
|
|
|
|
#include <linux/config.h>
|
|
#include <asm/atomic.h>
|
|
#include <asm/bitops.h>
|
|
#include <asm/mmu.h>
|
|
#include <asm/cputable.h>
|
|
|
|
/*
|
|
* On 32-bit PowerPC 6xx/7xx/7xxx CPUs, we use a set of 16 VSIDs
|
|
* (virtual segment identifiers) for each context. Although the
|
|
* hardware supports 24-bit VSIDs, and thus >1 million contexts,
|
|
* we only use 32,768 of them. That is ample, since there can be
|
|
* at most around 30,000 tasks in the system anyway, and it means
|
|
* that we can use a bitmap to indicate which contexts are in use.
|
|
* Using a bitmap means that we entirely avoid all of the problems
|
|
* that we used to have when the context number overflowed,
|
|
* particularly on SMP systems.
|
|
* -- paulus.
|
|
*/
|
|
|
|
/*
|
|
* This function defines the mapping from contexts to VSIDs (virtual
|
|
* segment IDs). We use a skew on both the context and the high 4 bits
|
|
* of the 32-bit virtual address (the "effective segment ID") in order
|
|
* to spread out the entries in the MMU hash table. Note, if this
|
|
* function is changed then arch/ppc/mm/hashtable.S will have to be
|
|
* changed to correspond.
|
|
*/
|
|
#define CTX_TO_VSID(ctx, va) (((ctx) * (897 * 16) + ((va) >> 28) * 0x111) \
|
|
& 0xffffff)
|
|
|
|
/*
|
|
The MPC8xx has only 16 contexts. We rotate through them on each
|
|
task switch. A better way would be to keep track of tasks that
|
|
own contexts, and implement an LRU usage. That way very active
|
|
tasks don't always have to pay the TLB reload overhead. The
|
|
kernel pages are mapped shared, so the kernel can run on behalf
|
|
of any task that makes a kernel entry. Shared does not mean they
|
|
are not protected, just that the ASID comparison is not performed.
|
|
-- Dan
|
|
|
|
The IBM4xx has 256 contexts, so we can just rotate through these
|
|
as a way of "switching" contexts. If the TID of the TLB is zero,
|
|
the PID/TID comparison is disabled, so we can use a TID of zero
|
|
to represent all kernel pages as shared among all contexts.
|
|
-- Dan
|
|
*/
|
|
|
|
static inline void enter_lazy_tlb(struct mm_struct *mm, struct task_struct *tsk)
|
|
{
|
|
}
|
|
|
|
#ifdef CONFIG_8xx
|
|
#define NO_CONTEXT 16
|
|
#define LAST_CONTEXT 15
|
|
#define FIRST_CONTEXT 0
|
|
|
|
#elif defined(CONFIG_4xx)
|
|
#define NO_CONTEXT 256
|
|
#define LAST_CONTEXT 255
|
|
#define FIRST_CONTEXT 1
|
|
|
|
#elif defined(CONFIG_E200) || defined(CONFIG_E500)
|
|
#define NO_CONTEXT 256
|
|
#define LAST_CONTEXT 255
|
|
#define FIRST_CONTEXT 1
|
|
|
|
#else
|
|
|
|
/* PPC 6xx, 7xx CPUs */
|
|
#define NO_CONTEXT ((mm_context_t) -1)
|
|
#define LAST_CONTEXT 32767
|
|
#define FIRST_CONTEXT 1
|
|
#endif
|
|
|
|
/*
|
|
* Set the current MMU context.
|
|
* On 32-bit PowerPCs (other than the 8xx embedded chips), this is done by
|
|
* loading up the segment registers for the user part of the address space.
|
|
*
|
|
* Since the PGD is immediately available, it is much faster to simply
|
|
* pass this along as a second parameter, which is required for 8xx and
|
|
* can be used for debugging on all processors (if you happen to have
|
|
* an Abatron).
|
|
*/
|
|
extern void set_context(mm_context_t context, pgd_t *pgd);
|
|
|
|
/*
|
|
* Bitmap of contexts in use.
|
|
* The size of this bitmap is LAST_CONTEXT + 1 bits.
|
|
*/
|
|
extern unsigned long context_map[];
|
|
|
|
/*
|
|
* This caches the next context number that we expect to be free.
|
|
* Its use is an optimization only, we can't rely on this context
|
|
* number to be free, but it usually will be.
|
|
*/
|
|
extern mm_context_t next_mmu_context;
|
|
|
|
/*
|
|
* If we don't have sufficient contexts to give one to every task
|
|
* that could be in the system, we need to be able to steal contexts.
|
|
* These variables support that.
|
|
*/
|
|
#if LAST_CONTEXT < 30000
|
|
#define FEW_CONTEXTS 1
|
|
extern atomic_t nr_free_contexts;
|
|
extern struct mm_struct *context_mm[LAST_CONTEXT+1];
|
|
extern void steal_context(void);
|
|
#endif
|
|
|
|
/*
|
|
* Get a new mmu context for the address space described by `mm'.
|
|
*/
|
|
static inline void get_mmu_context(struct mm_struct *mm)
|
|
{
|
|
mm_context_t ctx;
|
|
|
|
if (mm->context != NO_CONTEXT)
|
|
return;
|
|
#ifdef FEW_CONTEXTS
|
|
while (atomic_dec_if_positive(&nr_free_contexts) < 0)
|
|
steal_context();
|
|
#endif
|
|
ctx = next_mmu_context;
|
|
while (test_and_set_bit(ctx, context_map)) {
|
|
ctx = find_next_zero_bit(context_map, LAST_CONTEXT+1, ctx);
|
|
if (ctx > LAST_CONTEXT)
|
|
ctx = 0;
|
|
}
|
|
next_mmu_context = (ctx + 1) & LAST_CONTEXT;
|
|
mm->context = ctx;
|
|
#ifdef FEW_CONTEXTS
|
|
context_mm[ctx] = mm;
|
|
#endif
|
|
}
|
|
|
|
/*
|
|
* Set up the context for a new address space.
|
|
*/
|
|
#define init_new_context(tsk,mm) (((mm)->context = NO_CONTEXT), 0)
|
|
|
|
/*
|
|
* We're finished using the context for an address space.
|
|
*/
|
|
static inline void destroy_context(struct mm_struct *mm)
|
|
{
|
|
preempt_disable();
|
|
if (mm->context != NO_CONTEXT) {
|
|
clear_bit(mm->context, context_map);
|
|
mm->context = NO_CONTEXT;
|
|
#ifdef FEW_CONTEXTS
|
|
atomic_inc(&nr_free_contexts);
|
|
#endif
|
|
}
|
|
preempt_enable();
|
|
}
|
|
|
|
static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
|
|
struct task_struct *tsk)
|
|
{
|
|
#ifdef CONFIG_ALTIVEC
|
|
asm volatile (
|
|
BEGIN_FTR_SECTION
|
|
"dssall;\n"
|
|
#ifndef CONFIG_POWER4
|
|
"sync;\n" /* G4 needs a sync here, G5 apparently not */
|
|
#endif
|
|
END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
|
|
: : );
|
|
#endif /* CONFIG_ALTIVEC */
|
|
|
|
tsk->thread.pgdir = next->pgd;
|
|
|
|
/* No need to flush userspace segments if the mm doesnt change */
|
|
if (prev == next)
|
|
return;
|
|
|
|
/* Setup new userspace context */
|
|
get_mmu_context(next);
|
|
set_context(next->context, next->pgd);
|
|
}
|
|
|
|
#define deactivate_mm(tsk,mm) do { } while (0)
|
|
|
|
/*
|
|
* After we have set current->mm to a new value, this activates
|
|
* the context for the new mm so we see the new mappings.
|
|
*/
|
|
#define activate_mm(active_mm, mm) switch_mm(active_mm, mm, current)
|
|
|
|
extern void mmu_context_init(void);
|
|
|
|
#endif /* __PPC_MMU_CONTEXT_H */
|
|
#endif /* __KERNEL__ */
|