From: Kui-Feng Lee <sinquersw@gmail.com>
To: Martin KaFai Lau <martin.lau@linux.dev>, thinker.li@gmail.com
Cc: kuifeng@meta.com, bpf@vger.kernel.org, ast@kernel.org,
song@kernel.org, kernel-team@meta.com, andrii@kernel.org,
davemarchevsky@meta.com, dvernet@meta.com
Subject: Re: [RFC bpf-next v4 2/6] bpf: Extend PTR_TO_BTF_ID to handle pointers to scalar and array types.
Date: Fri, 2 Feb 2024 17:03:02 -0800 [thread overview]
Message-ID: <f2dc101d-74eb-4cef-82ab-7be5acf88772@gmail.com> (raw)
In-Reply-To: <190cf3fc-5c5e-4044-9cdc-4804ee49a03f@linux.dev>
On 2/2/24 16:52, Martin KaFai Lau wrote:
> On 2/2/24 2:05 PM, thinker.li@gmail.com wrote:
>> From: Kui-Feng Lee <thinker.li@gmail.com>
>>
>> The verifier calls btf_struct_access() to check the access for
>> PTR_TO_BTF_ID. btf_struct_access() supported only pointer to struct types
>> (including union). We add the support of scalar types and array types.
>>
>> btf_reloc_array_access() is responsible for relocating the access from
>> the
>> whole array to an element in the array. That means to adjust the offset
>> relatively to the start of an element and change the type to the type of
>> the element. With this relocation, we can check the access against the
>> element type instead of the array type itself.
>>
>> After relocation, the struct types, including union types, will continue
>> the loop of btf_struct_walk(). Other types are treated as scalar types,
>> including pointers, and return from btf_struct_access().
>
> Unless there is an immediate use case to support PTR_MAYBE_NULL on a
> non-struct pointer, I would suggest to separate the other pointer type
> support from the current PTR_MAYBE_NULL feature patchset. afaik, they
> are orthogonal.
Totally agree!
>
>>
>> Signed-off-by: Kui-Feng Lee <thinker.li@gmail.com>
>> ---
>> kernel/bpf/btf.c | 61 ++++++++++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 61 insertions(+)
>>
>> diff --git a/kernel/bpf/btf.c b/kernel/bpf/btf.c
>> index 0847035bba99..d3f94d04c69d 100644
>> --- a/kernel/bpf/btf.c
>> +++ b/kernel/bpf/btf.c
>> @@ -6590,6 +6590,61 @@ static int btf_struct_walk(struct
>> bpf_verifier_log *log, const struct btf *btf,
>> return -EINVAL;
>> }
>> +/* Relocate the access relatively to the beginning of an element in an
>> + * array.
>> + *
>> + * The offset is adjusted relatively to the beginning of the element
>> and the
>> + * type is adjusted to the type of the element.
>> + *
>> + * Return NULL for scalar, enum, and pointer type.
>> + * Return a btf_type pointer for struct and union.
>> + */
>> +static const struct btf_type *
>> +btf_reloc_array_access(struct bpf_verifier_log *log, const struct btf
>> *btf,
>> + const struct btf_type *t, int *off, int size)
>> +{
>> + const struct btf_type *rt, *elem_type;
>> + u32 rt_size, elem_id, total_nelems, rt_id, elem_size;
>> + u32 elem_idx;
>> +
>> + rt = __btf_resolve_size(btf, t, &rt_size, &elem_type, &elem_id,
>> + &total_nelems, &rt_id);
>> + if (IS_ERR(rt))
>> + return rt;
>> + if (btf_type_is_array(rt)) {
>> + if (*off >= rt_size) {
>> + bpf_log(log, "access out of range of type %s with offset
>> %d and size %u\n",
>> + __btf_name_by_offset(btf, t->name_off), *off, rt_size);
>> + return ERR_PTR(-EACCES);
>> + }
>> +
>> + /* Multi-dimensional arrays are flattened by
>> + * __btf_resolve_size(). Check the comment in
>> + * btf_struct_walk().
>> + */
>> + elem_size = rt_size / total_nelems;
>> + elem_idx = *off / elem_size;
>> + /* Relocate the offset relatively to the start of the
>> + * element at elem_idx.
>> + */
>> + *off -= elem_idx * elem_size;
>> + rt = elem_type;
>> + rt_size = elem_size;
>> + }
>> +
>> + if (btf_type_is_struct(rt))
>> + return rt;
>> +
>> + if (*off + size > rt_size) {
>> + bpf_log(log, "access beyond the range of type %s with offset
>> %d and size %d\n",
>> + __btf_name_by_offset(btf, rt->name_off), *off, size);
>> + return ERR_PTR(-EACCES);
>> + }
>> +
>> + /* The access is accepted as a scalar. */
>> + return NULL;
>> +}
>> +
>> int btf_struct_access(struct bpf_verifier_log *log,
>> const struct bpf_reg_state *reg,
>> int off, int size, enum bpf_access_type atype
>> __maybe_unused,
>> @@ -6625,6 +6680,12 @@ int btf_struct_access(struct bpf_verifier_log
>> *log,
>> }
>> t = btf_type_by_id(btf, id);
>> + t = btf_reloc_array_access(log, btf, t, &off, size);
>> + if (IS_ERR(t))
>> + return PTR_ERR(t);
>> + if (!t)
>> + return SCALAR_VALUE;
>> +
>> do {
>> err = btf_struct_walk(log, btf, t, off, size, &id,
>> &tmp_flag, field_name);
>
next prev parent reply other threads:[~2024-02-03 1:03 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-02 22:05 [RFC bpf-next v4 0/6] Support PTR_MAYBE_NULL for struct_ops arguments thinker.li
2024-02-02 22:05 ` [RFC bpf-next v4 1/6] bpf: Allow PTR_TO_BTF_ID even for pointers to int thinker.li
2024-02-02 22:05 ` [RFC bpf-next v4 2/6] bpf: Extend PTR_TO_BTF_ID to handle pointers to scalar and array types thinker.li
2024-02-03 0:52 ` Martin KaFai Lau
2024-02-03 1:03 ` Kui-Feng Lee [this message]
2024-02-02 22:05 ` [RFC bpf-next v4 3/6] bpf: Remove an unnecessary check thinker.li
2024-02-03 0:46 ` Martin KaFai Lau
2024-02-03 1:03 ` Kui-Feng Lee
2024-02-02 22:05 ` [RFC bpf-next v4 4/6] bpf: add btf pointer to struct bpf_ctx_arg_aux thinker.li
2024-02-02 22:05 ` [RFC bpf-next v4 5/6] bpf: Create argument information for nullable arguments thinker.li
2024-02-03 0:40 ` Martin KaFai Lau
2024-02-03 1:57 ` Kui-Feng Lee
2024-02-04 0:21 ` Kui-Feng Lee
2024-02-05 1:53 ` Kui-Feng Lee
2024-02-02 22:05 ` [RFC bpf-next v4 6/6] selftests/bpf: Test PTR_MAYBE_NULL arguments of struct_ops operators thinker.li
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=f2dc101d-74eb-4cef-82ab-7be5acf88772@gmail.com \
--to=sinquersw@gmail.com \
--cc=andrii@kernel.org \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=davemarchevsky@meta.com \
--cc=dvernet@meta.com \
--cc=kernel-team@meta.com \
--cc=kuifeng@meta.com \
--cc=martin.lau@linux.dev \
--cc=song@kernel.org \
--cc=thinker.li@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox