Re: [PATCH bpf-next v2] bpf: Check return from set_memory_rox() and friends

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




Le 15/03/2024 à 15:34, Daniel Borkmann a écrit :
> On 2/21/24 3:45 PM, Christophe Leroy wrote:
>> arch_protect_bpf_trampoline() and alloc_new_pack() call
>> set_memory_rox() which can fail, leading to unprotected memory.
>>
>> Take into account return from set_memory_XX() functions and add
>> __must_check flag to arch_protect_bpf_trampoline().
>>
>> Signed-off-by: Christophe Leroy <christophe.leroy@xxxxxxxxxx>
>> Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>
>> ---
>> v2:
>> - Move list_add_tail(&pack->list, &pack_list) at the end of 
>> alloc_new_pack()
>> - Split 2 lines that are reported longer than 80 chars by BPF 
>> patchwork's checkpatch report.
>> ---
>>   arch/x86/net/bpf_jit_comp.c    |  6 ++++--
>>   include/linux/bpf.h            |  4 ++--
>>   kernel/bpf/bpf_struct_ops.c    |  9 +++++++--
>>   kernel/bpf/core.c              | 29 ++++++++++++++++++++++-------
>>   kernel/bpf/trampoline.c        | 18 ++++++++++++------
>>   net/bpf/bpf_dummy_struct_ops.c |  4 +++-
>>   6 files changed, 50 insertions(+), 20 deletions(-)
>>
>> diff --git a/arch/x86/net/bpf_jit_comp.c b/arch/x86/net/bpf_jit_comp.c
>> index e1390d1e331b..128c8ec9164e 100644
>> --- a/arch/x86/net/bpf_jit_comp.c
>> +++ b/arch/x86/net/bpf_jit_comp.c
>> @@ -2780,12 +2780,14 @@ void arch_free_bpf_trampoline(void *image, 
>> unsigned int size)
>>       bpf_prog_pack_free(image, size);
>>   }
>> -void arch_protect_bpf_trampoline(void *image, unsigned int size)
>> +int arch_protect_bpf_trampoline(void *image, unsigned int size)
>>   {
>> +    return 0;
>>   }
>> -void arch_unprotect_bpf_trampoline(void *image, unsigned int size)
>> +int arch_unprotect_bpf_trampoline(void *image, unsigned int size)
>>   {
>> +    return 0;
>>   }
>>   int arch_prepare_bpf_trampoline(struct bpf_tramp_image *im, void 
>> *image, void *image_end,
>> diff --git a/include/linux/bpf.h b/include/linux/bpf.h
>> index b86bd15a051d..bb2723c264df 100644
>> --- a/include/linux/bpf.h
>> +++ b/include/linux/bpf.h
>> @@ -1116,8 +1116,8 @@ int arch_prepare_bpf_trampoline(struct 
>> bpf_tramp_image *im, void *image, void *i
>>                   void *func_addr);
>>   void *arch_alloc_bpf_trampoline(unsigned int size);
>>   void arch_free_bpf_trampoline(void *image, unsigned int size);
>> -void arch_protect_bpf_trampoline(void *image, unsigned int size);
>> -void arch_unprotect_bpf_trampoline(void *image, unsigned int size);
>> +int __must_check arch_protect_bpf_trampoline(void *image, unsigned 
>> int size);
>> +int arch_unprotect_bpf_trampoline(void *image, unsigned int size);
>>   int arch_bpf_trampoline_size(const struct btf_func_model *m, u32 flags,
>>                    struct bpf_tramp_links *tlinks, void *func_addr);
>> diff --git a/kernel/bpf/bpf_struct_ops.c b/kernel/bpf/bpf_struct_ops.c
>> index 0decd862dfe0..d920afb0dd60 100644
>> --- a/kernel/bpf/bpf_struct_ops.c
>> +++ b/kernel/bpf/bpf_struct_ops.c
>> @@ -488,7 +488,9 @@ static long bpf_struct_ops_map_update_elem(struct 
>> bpf_map *map, void *key,
>>               if (err)
>>                   goto reset_unlock;
>>           }
>> -        arch_protect_bpf_trampoline(st_map->image, PAGE_SIZE);
>> +        err = arch_protect_bpf_trampoline(st_map->image, PAGE_SIZE);
>> +        if (err)
>> +            goto reset_unlock;
>>           /* Let bpf_link handle registration & unregistration.
>>            *
>>            * Pair with smp_load_acquire() during lookup_elem().
>> @@ -497,7 +499,10 @@ static long bpf_struct_ops_map_update_elem(struct 
>> bpf_map *map, void *key,
>>           goto unlock;
>>       }
>> -    arch_protect_bpf_trampoline(st_map->image, PAGE_SIZE);
>> +    err = arch_protect_bpf_trampoline(st_map->image, PAGE_SIZE);
>> +    if (err)
>> +        goto reset_unlock;
>> +
>>       err = st_ops->reg(kdata);
>>       if (likely(!err)) {
>>           /* This refcnt increment on the map here after
>> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
>> index c49619ef55d0..eb2256ba6daf 100644
>> --- a/kernel/bpf/core.c
>> +++ b/kernel/bpf/core.c
>> @@ -898,23 +898,31 @@ static LIST_HEAD(pack_list);
>>   static struct bpf_prog_pack *alloc_new_pack(bpf_jit_fill_hole_t 
>> bpf_fill_ill_insns)
>>   {
>>       struct bpf_prog_pack *pack;
>> +    int err;
>>       pack = kzalloc(struct_size(pack, bitmap, 
>> BITS_TO_LONGS(BPF_PROG_CHUNK_COUNT)),
>>                  GFP_KERNEL);
>>       if (!pack)
>>           return NULL;
>>       pack->ptr = bpf_jit_alloc_exec(BPF_PROG_PACK_SIZE);
>> -    if (!pack->ptr) {
>> -        kfree(pack);
>> -        return NULL;
>> -    }
>> +    if (!pack->ptr)
>> +        goto out;
>>       bpf_fill_ill_insns(pack->ptr, BPF_PROG_PACK_SIZE);
>>       bitmap_zero(pack->bitmap, BPF_PROG_PACK_SIZE / 
>> BPF_PROG_CHUNK_SIZE);
>> -    list_add_tail(&pack->list, &pack_list);
>>       set_vm_flush_reset_perms(pack->ptr);
>> -    set_memory_rox((unsigned long)pack->ptr, BPF_PROG_PACK_SIZE / 
>> PAGE_SIZE);
>> +    err = set_memory_rox((unsigned long)pack->ptr,
>> +                 BPF_PROG_PACK_SIZE / PAGE_SIZE);
>> +    if (err)
>> +        goto out_free;
>> +    list_add_tail(&pack->list, &pack_list);
>>       return pack;
>> +
>> +out_free:
>> +    bpf_jit_free_exec(pack->ptr);
>> +out:
>> +    kfree(pack);
>> +    return NULL;
>>   }
>>   void *bpf_prog_pack_alloc(u32 size, bpf_jit_fill_hole_t 
>> bpf_fill_ill_insns)
>> @@ -929,9 +937,16 @@ void *bpf_prog_pack_alloc(u32 size, 
>> bpf_jit_fill_hole_t bpf_fill_ill_insns)
>>           size = round_up(size, PAGE_SIZE);
>>           ptr = bpf_jit_alloc_exec(size);
>>           if (ptr) {
>> +            int err;
>> +
>>               bpf_fill_ill_insns(ptr, size);
>>               set_vm_flush_reset_perms(ptr);
>> -            set_memory_rox((unsigned long)ptr, size / PAGE_SIZE);
>> +            err = set_memory_rox((unsigned long)ptr,
>> +                         size / PAGE_SIZE);
>> +            if (err) {
>> +                bpf_jit_free_exec(ptr);
>> +                ptr = NULL;
>> +            }
>>           }
>>           goto out;
>>       }
>> diff --git a/kernel/bpf/trampoline.c b/kernel/bpf/trampoline.c
>> index d382f5ebe06c..6e64ac9083b6 100644
>> --- a/kernel/bpf/trampoline.c
>> +++ b/kernel/bpf/trampoline.c
>> @@ -456,7 +456,9 @@ static int bpf_trampoline_update(struct 
>> bpf_trampoline *tr, bool lock_direct_mut
>>       if (err < 0)
>>           goto out_free;
>> -    arch_protect_bpf_trampoline(im->image, im->size);
>> +    err = arch_protect_bpf_trampoline(im->image, im->size);
>> +    if (err)
>> +        goto out_free;
>>       WARN_ON(tr->cur_image && total == 0);
>>       if (tr->cur_image)
>> @@ -1072,17 +1074,21 @@ void __weak arch_free_bpf_trampoline(void 
>> *image, unsigned int size)
>>       bpf_jit_free_exec(image);
>>   }
>> -void __weak arch_protect_bpf_trampoline(void *image, unsigned int size)
>> +int __weak arch_protect_bpf_trampoline(void *image, unsigned int size)
> 
> nit: Should we add __must_check as well here?

Don't think so.

Looks like only prototypes get the __must_check

See for instance device_create_bin_file(), there's the __must_check on 
the prototype in include/linux/device.h but not the definition in 
drivers/base/core.c

> 
>>   {
>>       WARN_ON_ONCE(size > PAGE_SIZE);
>> -    set_memory_rox((long)image, 1);
>> +    return set_memory_rox((long)image, 1);
>>   }
>> -void __weak arch_unprotect_bpf_trampoline(void *image, unsigned int 
>> size)
>> +int __weak arch_unprotect_bpf_trampoline(void *image, unsigned int size)
>>   {
>> +    int err;
>>       WARN_ON_ONCE(size > PAGE_SIZE);
>> -    set_memory_nx((long)image, 1);
>> -    set_memory_rw((long)image, 1);
>> +
>> +    err = set_memory_nx((long)image, 1);
>> +    if (err)
>> +        return err;
>> +    return set_memory_rw((long)image, 1);
>>   }
> 
> Do we still need this? It looks like this does not have an in-tree user 
> anymore.

Looks like last user went away with commit 187e2af05abe ("bpf: 
struct_ops supports more than one page for trampolines.") but I'm having 
hard time figuring if it's valid or not.

But as there is no user anymore it surely can go away. Will you drop it 
or do you want a proper patch from me ?


Christophe




[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux