Re: [PATCH v6 1/4] mm/slub: enable debugging memory wasting of kmalloc

[Vlastimil Babka <vbabka@suse.cz>]

On 11/1/22 11:33, John Thomson wrote:
> On Tue, 1 Nov 2022, at 09:31, Hyeonggon Yoo wrote:
>> On Tue, Nov 01, 2022 at 09:20:21AM +0000, John Thomson wrote:
>>> On Tue, 1 Nov 2022, at 07:57, Feng Tang wrote:
>>> > Hi Thomson,
>>> >
>>> > Thanks for testing!
>>> >
>>> > + mips maintainer and mail list. The original report is here
>>> >
>>> > https://lore.kernel.org/lkml/becf2ac3-2a90-4f3a-96d9-a70f67c66e4a@app.fastmail.com/
>>>
>>> I am guessing my issue comes from __kmem_cache_alloc_lru accessing s->object_size when (kmem_cache) s is NULL?
>>> If that is the case, this change is not to blame, it only exposes the issue?
>>> 
>>> I get the following dmesg (note very early NULL kmem_cache) with the below change atop v6.1-rc3:
>>> 
>>> transfer started ......................................... transfer ok, time=2.02s
>>> setting up elf image... OK
>>> jumping to kernel code
>>> zimage at:     80B842A0 810B4EFC
>>> 
>>> Uncompressing Linux at load address 80001000
>>> 
>>> Copy device tree to address  80B80EE0
>>> 
>>> Now, booting the kernel...
>>> 
>>> [    0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #61 SMP Tue Nov  1 18:04:13 AEST 2022
>>> [    0.000000] slub: kmem_cache_alloc called with kmem_cache: 0x0
>>> [    0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache: 0x0
>>> [    0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3
>>> [    0.000000] printk: bootconsole [early0] enabled
>>> [    0.000000] CPU0 revision is: 0001992f (MIPS 1004Kc)
>>> [    0.000000] MIPS: machine is MikroTik RouterBOARD 760iGS
>>> 
>>> normal boot
>>> 
>>> 
>>> diff --git a/mm/slub.c b/mm/slub.c
>>> index 157527d7101b..10fcdf2520d2 100644
>>> --- a/mm/slub.c
>>> +++ b/mm/slub.c
>>> @@ -3410,7 +3410,13 @@ static __always_inline
>>>  void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>>  			     gfp_t gfpflags)
>>>  {
>>> -	void *ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
>>> +	void *ret;
>>> +	if (IS_ERR_OR_NULL(s)) {
>>> +		pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
>>> +		ret = slab_alloc(s, lru, gfpflags, _RET_IP_, 0);
>>> +	} else {
>>> +		ret = slab_alloc(s, lru, gfpflags, _RET_IP_, s->object_size);
>>> +	}
>>>  
>>>  	trace_kmem_cache_alloc(_RET_IP_, ret, s, gfpflags, NUMA_NO_NODE);
>>>  
>>> @@ -3419,6 +3425,8 @@ void *__kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>>  
>>>  void *kmem_cache_alloc(struct kmem_cache *s, gfp_t gfpflags)
>>>  {
>>> +	if (IS_ERR_OR_NULL(s))
>>> +		pr_warn("slub: kmem_cache_alloc called with kmem_cache: %pSR\n", s);
>>>  	return __kmem_cache_alloc_lru(s, NULL, gfpflags);
>>>  }
>>>  EXPORT_SYMBOL(kmem_cache_alloc);
>>> @@ -3426,6 +3434,8 @@ EXPORT_SYMBOL(kmem_cache_alloc);
>>>  void *kmem_cache_alloc_lru(struct kmem_cache *s, struct list_lru *lru,
>>>  			   gfp_t gfpflags)
>>>  {
>>> +	if (IS_ERR_OR_NULL(s))
>>> +		pr_warn("slub: __kmem_cache_alloc_lru called with kmem_cache: %pSR\n", s);
>>>  	return __kmem_cache_alloc_lru(s, lru, gfpflags);
>>>  }
>>>  EXPORT_SYMBOL(kmem_cache_alloc_lru);
>>> 
>>> 
>>> Any hints on where kmem_cache_alloc would be being called from this early?
>>> I will start looking from /init/main.c around pr_notice("%s", linux_banner);
>>
>> Great. Would you try calling dump_stack(); when we observed s == NULL?
>> That would give more information about who passed s == NULL to these
>> functions.
>>
> 
> With the dump_stack() in place:
> 
> Now, booting the kernel...
> 
> [    0.000000] Linux version 6.1.0-rc3+ (john@john) (mipsel-buildroot-linux-gnu-gcc.br_real (Buildroot 2021.11-4428-g6b6741b) 12.2.0, GNU ld (GNU Binutils) 2.39) #62 SMP Tue Nov  1 19:49:52 AEST 2022
> [    0.000000] slub: __kmem_cache_alloc_lru called with kmem_cache ptr: 0x0
> [    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.1.0-rc3+ #62
> [    0.000000] Stack : 810fff78 80084d98 80889d00 00000004 00000000 00000000 80889d5c 80c90000
> [    0.000000]         80920000 807bd380 8089d368 80923bd3 00000000 00000001 80889d08 00000000
> [    0.000000]         00000000 00000000 807bd380 8084bd51 00000002 00000002 00000001 6d6f4320
> [    0.000000]         00000000 80c97ce9 80c97d14 fffffffc 807bd380 00000000 00000003 00000dc0
> [    0.000000]         00000000 a0000000 80910000 8110a0b4 00000000 00000020 80010000 80010000
> [    0.000000]         ...
> [    0.000000] Call Trace:
> [    0.000000] [<80008260>] show_stack+0x28/0xf0
> [    0.000000] [<8070cdc0>] dump_stack_lvl+0x60/0x80
> [    0.000000] [<801c1428>] kmem_cache_alloc+0x5c0/0x740
> [    0.000000] [<8092856c>] prom_soc_init+0x1fc/0x2b4
> [    0.000000] [<80928060>] prom_init+0x44/0xf0
> [    0.000000] [<80929214>] setup_arch+0x4c/0x6a8
> [    0.000000] [<809257e0>] start_kernel+0x88/0x7c0
> [    0.000000] 
> [    0.000000] SoC Type: MediaTek MT7621 ver:1 eco:3

The stack means CONFIG_TRACING=n, is that right?

That would mean
prom_soc_init()
  soc_dev_init()
    kzalloc() -> kmalloc()
      kmalloc_trace()  // after #else /* CONFIG_TRACING */
        kmem_cache_alloc(s, flags);

Looks like this path is a small bug in the wasting detection patch, as we
throw away size there.

AFAICS before this patch, we "survive" "kmem_cache *s" being NULL as
slab_pre_alloc_hook() will happen to return NULL and we bail out from
slab_alloc_node(). But this is a side-effect, not an intended protection.
Also the CONFIG_TRACING variant of kmalloc_trace() would have called
trace_kmalloc dereferencing s->size anyway even before this patch.

I don't think we should add WARNS in the slab hot paths just to prevent this
rare error of using slab too early. At most VM_WARN... would be acceptable
but still not necessary as crashing immediately from a NULL pointer is
sufficient.

So IMHO mips should fix their soc init, and we should look into the
CONFIG_TRACING=n variant of kmalloc_trace(), to pass orig_size properly.