From: Adhemerval Zanella <adhemerval.zanella@linaro.org>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: GNU C Library <libc-alpha@sourceware.org>
Subject: Re: [PATCH 1/2] Initial support for GNU_PROPERTY_1_NEEDED
Date: Thu, 7 Oct 2021 13:53:14 -0300 [thread overview]
Message-ID: <e1de4793-75ac-4b36-7bc0-9bb20efdb7c9@linaro.org> (raw)
In-Reply-To: <CAMe9rOr-dVEbrEwQeud5oXFi3eLcpnGM7giE=uxgNrmt817a6g@mail.gmail.com>
On 07/10/2021 12:11, H.J. Lu wrote:
> On Thu, Oct 7, 2021 at 7:53 AM Adhemerval Zanella
> <adhemerval.zanella@linaro.org> wrote:
>>
>>
>>
>> On 03/08/2021 18:59, H.J. Lu via Libc-alpha wrote:
>>> 1. Add GNU_PROPERTY_1_NEEDED:
>>>
>>> #define GNU_PROPERTY_1_NEEDED GNU_PROPERTY_UINT32_OR_LO
>>>
>>> to indicate the needed properties by the object file.
>>> 2. Add GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS:
>>>
>>> #define GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS (1U << 0)
>>>
>>> to indicate that the object file requires canonical function pointers and
>>> cannot be used with copy relocation.
>>> 3. Scan GNU_PROPERTY_1_NEEDED property and store it in l_1_needed.
>>
>> It was added by binutils with commit 6320fd00dc374f74658, maybe add on
>> the commit the version expected to have this support.
>>
>> LGTM, with a small suggestion below.
>>
>> Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
>>
>>> ---
>>> elf/elf.h | 17 +++++++++++++++++
>>> sysdeps/generic/dl-prop.h | 9 ++++++++-
>>> sysdeps/generic/link_map.h | 3 ++-
>>> sysdeps/x86/dl-prop.h | 19 ++++++++++++++-----
>>> sysdeps/x86/link_map.h | 2 ++
>>> 5 files changed, 43 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/elf/elf.h b/elf/elf.h
>>> index 4738dfa28f..50f87baceb 100644
>>> --- a/elf/elf.h
>>> +++ b/elf/elf.h
>>> @@ -1312,6 +1312,23 @@ typedef struct
>>> /* No copy relocation on protected data symbol. */
>>> #define GNU_PROPERTY_NO_COPY_ON_PROTECTED 2
>>>
>>> +/* A 4-byte unsigned integer property: A bit is set if it is set in all
>>> + relocatable inputs. */
>>> +#define GNU_PROPERTY_UINT32_AND_LO 0xb0000000
>>> +#define GNU_PROPERTY_UINT32_AND_HI 0xb0007fff
>>> +
>>> +/* A 4-byte unsigned integer property: A bit is set if it is set in any
>>> + relocatable inputs. */
>>> +#define GNU_PROPERTY_UINT32_OR_LO 0xb0008000
>>> +#define GNU_PROPERTY_UINT32_OR_HI 0xb000ffff
>>> +
>>> +/* The needed properties by the object file. */
>>> +#define GNU_PROPERTY_1_NEEDED GNU_PROPERTY_UINT32_OR_LO
>>> +
>>> +/* Set if the object file requires canonical function pointers and
>>> + cannot be used with copy relocation. */
>>> +#define GNU_PROPERTY_1_NEEDED_INDIRECT_EXTERN_ACCESS (1U << 0)
>>> +
>>> /* Processor-specific semantics, lo */
>>> #define GNU_PROPERTY_LOPROC 0xc0000000
>>> /* Processor-specific semantics, hi */
>>
>> Ok.
>>
>>> diff --git a/sysdeps/generic/dl-prop.h b/sysdeps/generic/dl-prop.h
>>> index eaee8052b6..207aadb35d 100644
>>> --- a/sysdeps/generic/dl-prop.h
>>> +++ b/sysdeps/generic/dl-prop.h
>>> @@ -47,7 +47,14 @@ static inline int __attribute__ ((always_inline))
>>> _dl_process_gnu_property (struct link_map *l, int fd, uint32_t type,
>>> uint32_t datasz, void *data)
>>> {
>>> - return 0;
>>> + /* Continue until GNU_PROPERTY_1_NEEDED is found. */
>>> + if (type == GNU_PROPERTY_1_NEEDED)
>>> + {
>>> + if (datasz == 4)
>>> + l->l_1_needed = *(unsigned int *) data;
>>
>> Although the pointer seems to be aligned, I think it would be better to
>> use memcpy here:
>>
>> unsigned int t;
>> memcpy (&t, data, sizeof (t));
>> l->l_1_needed = t;
>>
>> (the x86 header might use a more relax since the architecture always
>> support unaligned access)
>
> The GNU property is guaranteed to be properly aligned:
>
> void
> _dl_process_pt_gnu_property (struct link_map *l, int fd, const ElfW(Phdr) *ph)
> {
> const ElfW(Nhdr) *note = (const void *) (ph->p_vaddr + l->l_addr);
> const ElfW(Addr) size = ph->p_memsz;
> const ElfW(Addr) align = ph->p_align;
>
> /* The NT_GNU_PROPERTY_TYPE_0 note must be aligned to 4 bytes in
> 32-bit objects and to 8 bytes in 64-bit objects. Skip notes
> with incorrect alignment. */
> if (align != (__ELF_NATIVE_CLASS / 8))
> return;
>
> data passed to _dl_process_gnu_property should be properly aligned.
Alright, the memcpy seems unnecessary then.
next prev parent reply other threads:[~2021-10-07 16:53 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-03 21:59 H.J. Lu
2021-08-03 21:59 ` [PATCH 2/2] Add run-time chesk for indirect external access H.J. Lu
2021-09-27 2:10 ` PING^1 " H.J. Lu
2021-10-07 16:58 ` Adhemerval Zanella
2021-10-07 17:10 ` H.J. Lu
2021-10-07 17:45 ` Adhemerval Zanella
2021-10-07 14:53 ` [PATCH 1/2] Initial support for GNU_PROPERTY_1_NEEDED Adhemerval Zanella
2021-10-07 15:11 ` H.J. Lu
2021-10-07 16:53 ` Adhemerval Zanella [this message]
2021-10-07 17:04 ` H.J. Lu
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=e1de4793-75ac-4b36-7bc0-9bb20efdb7c9@linaro.org \
--to=adhemerval.zanella@linaro.org \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@sourceware.org \
/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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).