From: Kyrill Tkachov <kyrylo.tkachov@foss.arm.com>
To: Christophe Lyon <christophe.lyon@st.com>,
Christophe Lyon <christophe.lyon@linaro.org>
Cc: "gcc-patches@gcc.gnu.org" <gcc-patches@gcc.gnu.org>
Subject: Re: [ARM/FDPIC v5 09/21] [ARM] FDPIC: Add support for taking address of nested function
Date: Fri, 30 Aug 2019 08:54:00 -0000 [thread overview]
Message-ID: <7bad7519-da2b-cf7b-a3be-15c86e6fe27e@foss.arm.com> (raw)
In-Reply-To: <2a92502b-c4ea-d159-fb1b-e128b7c953fe@st.com>
On 8/29/19 4:36 PM, Christophe Lyon wrote:
> On 31/07/2019 16:44, Christophe Lyon wrote:
>> On Tue, 16 Jul 2019 at 14:42, Kyrill Tkachov
>> <kyrylo.tkachov@foss.arm.com> wrote:
>>>
>>>
>>> On 7/16/19 12:18 PM, Kyrill Tkachov wrote:
>>>> Hi Christophe
>>>>
>>>> On 5/15/19 1:39 PM, Christophe Lyon wrote:
>>>>> In FDPIC mode, the trampoline generated to support pointers to nested
>>>>> functions looks like:
>>>>>
>>>>> Â Â Â Â Â Â Â Â Â Â Â .word trampoline address
>>>>> Â Â Â Â Â Â Â Â Â Â Â .word trampoline GOT address
>>>>>            ldr           r12, [pc, #8]
>>>>>            ldr           r9, [pc, #8]
>>>>>            ldr          pc, [pc, #8]
>>>>> Â Â Â Â Â Â Â Â Â Â Â .word static chain value
>>>>> Â Â Â Â Â Â Â Â Â Â Â .word GOT address
>>>>> Â Â Â Â Â Â Â Â Â Â Â .word function's address
>>>>>
>>>>> because in FDPIC function pointers are actually pointers to function
>>>>> descriptors, we have to actually generate a function descriptor for
>>>>> the trampoline.
>>>>>
>>>>> 2019-XX-XXÂ Christophe Lyon <christophe.lyon@st.com>
>>>>>         Mickaël Guêné <mickael.guene@st.com>
>>>>>
>>>>> Â Â Â Â Â Â Â Â gcc/
>>>>> Â Â Â Â Â Â Â Â * config/arm/arm.c (arm_asm_trampoline_template): Add FDPIC
>>>>> Â Â Â Â Â Â Â Â support.
>>>>> Â Â Â Â Â Â Â Â (arm_trampoline_init): Likewise.
>>>>> Â Â Â Â Â Â Â Â (arm_trampoline_init): Likewise.
>>>>> Â Â Â Â Â Â Â Â * config/arm/arm.h (TRAMPOLINE_SIZE): Likewise.
>>>>>
>>>>> Change-Id: Idc4d5f629ae4f8d79bdf9623517481d524a0c144
>>>>>
>>>>> diff --git a/gcc/config/arm/arm.c b/gcc/config/arm/arm.c
>>>>> index 40e3f3b..99d13bf 100644
>>>>> --- a/gcc/config/arm/arm.c
>>>>> +++ b/gcc/config/arm/arm.c
>>>>> @@ -3976,13 +3976,50 @@ arm_warn_func_return (tree decl)
>>>>> Â Â Â Â Â Â Â Â Â Â Â Â .word static chain value
>>>>> Â Â Â Â Â Â Â Â Â Â Â Â .word function's address
>>>>> Â Â Â Â XXX FIXME: When the trampoline returns, r8 will be
>>>>> clobbered. */
>>>>> +/* In FDPIC mode, the trampoline looks like:
>>>>> +Â Â Â Â Â Â Â Â Â .word trampoline address
>>>>> +Â Â Â Â Â Â Â Â Â .word trampoline GOT address
>>>>> +         ldr           r12, [pc, #8] ; #4 for Thumb2
>>>>> +         ldr           r9, [pc, #8] ; #4 for Thumb2
>>>>> +         ldr          pc, [pc, #8] ; #4 for Thumb2
>>>>> +Â Â Â Â Â Â Â Â Â .word static chain value
>>>>> +Â Â Â Â Â Â Â Â Â .word GOT address
>>>>> +Â Â Â Â Â Â Â Â Â .word function's address
>>>>> +*/
>>>>
>>>>
>>>> I think this comment is not right for Thumb2.
>>>>
>>>> These load instructionshave 32-bit encodings, even in Thumb2 (they use
>>>> high registers).
>>>
>>> Andre and Wilco pointed out to me offline that the offset should be #4
>>> for Arm mode.
>>>
>>> The Arm ARM at E1.2.3 says:
>>>
>>> PC, the program counter
>>>
>>> * When executing an A32 instruction, PC reads as the address of the
>>> current instruction plus 8.
>>>
>>> * When executing a T32 instruction, PC reads as the address of the
>>> current instruction plus 4.
>>>
>>
>> Yes, it looks like the code is right, and the comment is wrong:
>> - offset 8 for thumb2 mode
>> - offset 4 for arm mode
>>
> Here is the updated version
>
Ok with a fixed ChangeLog (it currently mentions arm_trampoline_init
twice but doesn't mention arm_trampoline_adjust_address)
Thanks,
Kyrill
>> Thanks,
>>
>> Christophe
>>
>>> Thanks,
>>>
>>> Kyrill
>>>
>>>
>>>>
>>>> Also, please merge this comment with the one above (no separate /**/)
>>>>
>>>>>
>>>>> Â static void
>>>>> Â arm_asm_trampoline_template (FILE *f)
>>>>> Â {
>>>>> Â Â Â fprintf (f, "\t.syntax unified\n");
>>>>>
>>>>> -Â if (TARGET_ARM)
>>>>> +Â if (TARGET_FDPIC)
>>>>> +Â Â Â {
>>>>> +Â Â Â Â Â /* The first two words are a function descriptor pointing
>>>>> to the
>>>>> +       trampoline code just below. */
>>>>> +Â Â Â Â Â if (TARGET_ARM)
>>>>> +Â Â Â Â Â Â fprintf (f, "\t.arm\n");
>>>>> +Â Â Â Â Â else if (TARGET_THUMB2)
>>>>> +Â Â Â Â Â Â fprintf (f, "\t.thumb\n");
>>>>> +Â Â Â Â Â else
>>>>> +      /* Only ARM and Thumb-2 are supported. */
>>>>> +Â Â Â Â Â Â gcc_unreachable ();
>>>>> +
>>>>> +Â Â Â Â Â assemble_aligned_integer (UNITS_PER_WORD, const0_rtx);
>>>>> +Â Â Â Â Â assemble_aligned_integer (UNITS_PER_WORD, const0_rtx);
>>>>> +Â Â Â Â Â /* Trampoline code which sets the static chain register but
>>>>> also
>>>>> +Â Â Â Â Â Â Â PIC register before jumping into real code. */
>>>>> +Â Â Â Â Â asm_fprintf (f, "\tldr\t%r, [%r, #%d]\n",
>>>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â STATIC_CHAIN_REGNUM, PC_REGNUM,
>>>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â TARGET_THUMB2 ? 8 : 4);
>>>>> +Â Â Â Â Â asm_fprintf (f, "\tldr\t%r, [%r, #%d]\n",
>>>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â PIC_OFFSET_TABLE_REGNUM, PC_REGNUM,
>>>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â TARGET_THUMB2 ? 8 : 4);
>>>>> +Â Â Â Â Â asm_fprintf (f, "\tldr\t%r, [%r, #%d]\n",
>>>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â PC_REGNUM, PC_REGNUM,
>>>>> +Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â TARGET_THUMB2 ? 8 : 4);
>>>>
>>>>
>>>> As above, I think the offset should be 8 for both Arm and Thumb2.
>>>>
>>>> Thanks,
>>>>
>>>> Kyrill
>>>>
>>>>
>>>>> +Â Â Â Â Â assemble_aligned_integer (UNITS_PER_WORD, const0_rtx);
>>>>> +Â Â Â }
>>>>> +Â else if (TARGET_ARM)
>>>>> Â Â Â Â Â {
>>>>> Â Â Â Â Â Â Â fprintf (f, "\t.arm\n");
>>>>> Â Â Â Â Â Â Â asm_fprintf (f, "\tldr\t%r, [%r, #0]\n", STATIC_CHAIN_REGNUM,
>>>>> PC_REGNUM);
>>>>> @@ -4023,12 +4060,40 @@ arm_trampoline_init (rtx m_tramp, tree
>>>>> fndecl,
>>>>> rtx chain_value)
>>>>> Â Â Â emit_block_move (m_tramp, assemble_trampoline_template (),
>>>>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â GEN_INT (TRAMPOLINE_SIZE), BLOCK_OP_NORMAL);
>>>>>
>>>>> -Â mem = adjust_address (m_tramp, SImode, TARGET_32BIT ? 8 : 12);
>>>>> -Â emit_move_insn (mem, chain_value);
>>>>> +Â if (TARGET_FDPIC)
>>>>> +Â Â Â {
>>>>> +Â Â Â Â Â rtx funcdesc = XEXP (DECL_RTL (fndecl), 0);
>>>>> +Â Â Â Â Â rtx fnaddr = gen_rtx_MEM (Pmode, funcdesc);
>>>>> +Â Â Â Â Â rtx gotaddr = gen_rtx_MEM (Pmode, plus_constant (Pmode,
>>>>> funcdesc, 4));
>>>>> +Â Â Â Â Â /* The function start address is at offset 8, but in Thumb
>>>>> mode
>>>>> +Â Â Â Â Â Â Â we want bit 0 set to 1 to indicate Thumb-ness, hence 9
>>>>> +       below. */
>>>>> +Â Â Â Â Â rtx trampoline_code_start
>>>>> +Â Â Â Â Â Â = plus_constant (Pmode, XEXP (m_tramp, 0), TARGET_THUMB2 ? 9
>>>> : 8);
>>>>> +
>>>>> +     /* Write initial funcdesc which points to the trampoline. */
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, 0);
>>>>> +Â Â Â Â Â emit_move_insn (mem, trampoline_code_start);
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, 4);
>>>>> +Â Â Â Â Â emit_move_insn (mem, gen_rtx_REG (Pmode,
>>>> PIC_OFFSET_TABLE_REGNUM));
>>>>> +     /* Setup static chain. */
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, 20);
>>>>> +Â Â Â Â Â emit_move_insn (mem, chain_value);
>>>>> +     /* GOT + real function entry point. */
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, 24);
>>>>> +Â Â Â Â Â emit_move_insn (mem, gotaddr);
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, 28);
>>>>> +Â Â Â Â Â emit_move_insn (mem, fnaddr);
>>>>> +Â Â Â }
>>>>> +Â else
>>>>> +Â Â Â {
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, TARGET_32BIT ? 8 : 12);
>>>>> +Â Â Â Â Â emit_move_insn (mem, chain_value);
>>>>>
>>>>> -Â mem = adjust_address (m_tramp, SImode, TARGET_32BIT ? 12 : 16);
>>>>> -Â fnaddr = XEXP (DECL_RTL (fndecl), 0);
>>>>> -Â emit_move_insn (mem, fnaddr);
>>>>> +Â Â Â Â Â mem = adjust_address (m_tramp, SImode, TARGET_32BIT ? 12 :
>>>>> 16);
>>>>> +Â Â Â Â Â fnaddr = XEXP (DECL_RTL (fndecl), 0);
>>>>> +Â Â Â Â Â emit_move_insn (mem, fnaddr);
>>>>> +Â Â Â }
>>>>>
>>>>> Â Â Â a_tramp = XEXP (m_tramp, 0);
>>>>> Â Â Â emit_library_call (gen_rtx_SYMBOL_REF (Pmode, "__clear_cache"),
>>>>> @@ -4042,7 +4107,9 @@ arm_trampoline_init (rtx m_tramp, tree fndecl,
>>>>> rtx chain_value)
>>>>> Â static rtx
>>>>> Â arm_trampoline_adjust_address (rtx addr)
>>>>> Â {
>>>>> -Â if (TARGET_THUMB)
>>>>> +Â /* For FDPIC don't fix trampoline address since it's a function
>>>>> +    descriptor and not a function address. */
>>>>> +Â if (TARGET_THUMB && !TARGET_FDPIC)
>>>>> Â Â Â Â Â addr = expand_simple_binop (Pmode, IOR, addr, const1_rtx,
>>>>> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â NULL, 0, OPTAB_LIB_WIDEN);
>>>>> Â Â Â return addr;
>>>>> diff --git a/gcc/config/arm/arm.h b/gcc/config/arm/arm.h
>>>>> index 45c0e2b..f80df63 100644
>>>>> --- a/gcc/config/arm/arm.h
>>>>> +++ b/gcc/config/arm/arm.h
>>>>> @@ -1602,7 +1602,7 @@ typedef struct
>>>>> Â #define INIT_EXPANDERSÂ arm_init_expanders ()
>>>>>
>>>>> Â /* Length in units of the trampoline for entering a nested
>>>> function. */
>>>>> -#define TRAMPOLINE_SIZE (TARGET_32BIT ? 16 : 20)
>>>>> +#define TRAMPOLINE_SIZEÂ (TARGET_FDPIC ? 32 : (TARGET_32BIT ? 16
>>>>> : 20))
>>>>>
>>>>>  /* Alignment required for a trampoline in bits. */
>>>>> Â #define TRAMPOLINE_ALIGNMENTÂ 32
>>>>> --
>>>>> 2.6.3
>>>>>
>> .
>>
>
next prev parent reply other threads:[~2019-08-30 8:32 UTC|newest]
Thread overview: 109+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-05-15 12:40 [ARM/FDPIC v5 00/21] FDPIC ABI for ARM Christophe Lyon
2019-05-15 12:40 ` [ARM/FDPIC v5 01/21] [ARM] FDPIC: Add -mfdpic option support Christophe Lyon
2019-07-16 10:18 ` Richard Sandiford
2019-08-29 15:08 ` Christophe Lyon
2019-08-30 10:06 ` Richard Sandiford
2019-05-15 12:41 ` [ARM/FDPIC v5 05/21] [ARM] FDPIC: Fix __do_global_dtors_aux and frame_dummy generation Christophe Lyon
2019-07-12 6:49 ` Richard Sandiford
2019-07-12 14:25 ` Christophe Lyon
2019-08-29 15:39 ` Christophe Lyon
2019-08-30 8:41 ` Richard Sandiford
2019-05-15 12:41 ` [ARM/FDPIC v5 02/21] [ARM] FDPIC: Handle arm*-*-uclinuxfdpiceabi in configure scripts Christophe Lyon
2019-07-12 7:44 ` Richard Sandiford
2019-07-12 13:25 ` Christophe Lyon
2019-07-12 13:28 ` Richard Sandiford
2019-08-29 15:14 ` Christophe Lyon
2019-08-30 9:30 ` Richard Sandiford
2019-08-30 14:48 ` Christophe Lyon
2019-08-30 15:15 ` Richard Sandiford
2019-09-02 7:51 ` Christophe Lyon
2019-09-02 8:31 ` Richard Sandiford
2019-08-30 14:40 ` Jonathan Wakely
2019-05-15 12:41 ` [ARM/FDPIC v5 03/21] [ARM] FDPIC: Force FDPIC related options unless -mno-fdpic is provided Christophe Lyon
2019-05-15 13:55 ` Szabolcs Nagy
2019-05-15 14:37 ` Rich Felker
2019-05-15 15:12 ` Christophe Lyon
2019-05-15 15:37 ` Rich Felker
2019-05-15 15:59 ` Szabolcs Nagy
2019-05-15 16:07 ` Rich Felker
2019-05-21 15:29 ` Christophe Lyon
2019-05-21 15:48 ` Rich Felker
2019-05-22 8:39 ` Szabolcs Nagy
2019-05-22 8:45 ` Christophe Lyon
2019-05-23 12:45 ` Christophe Lyon
2019-07-16 10:38 ` Richard Sandiford
2019-07-16 20:00 ` Rich Felker
2019-08-01 10:13 ` Christophe Lyon
2019-08-06 14:28 ` Richard Sandiford
2019-08-29 15:14 ` Christophe Lyon
2019-08-30 9:40 ` Richard Sandiford
2019-05-15 12:41 ` [ARM/FDPIC v5 04/21] [ARM] FDPIC: Add support for FDPIC for arm architecture Christophe Lyon
2019-07-16 12:33 ` Richard Sandiford
2019-08-20 17:13 ` Christophe Lyon
2019-08-29 15:39 ` Christophe Lyon
2019-09-02 16:12 ` Richard Sandiford
2019-09-02 20:04 ` Christophe Lyon
2019-09-03 8:40 ` Richard Sandiford
2019-09-04 19:59 ` Christophe Lyon
2019-09-05 8:03 ` Richard Sandiford
2019-05-15 12:42 ` [ARM/FDPIC v5 07/21] [ARM] FDPIC: Avoid saving/restoring r9 on stack since it is read-only Christophe Lyon
2019-07-16 10:42 ` Kyrill Tkachov
2019-05-15 12:42 ` [ARM/FDPIC v5 06/21] [ARM] FDPIC: Add support for c++ exceptions Christophe Lyon
2019-08-30 9:31 ` Kyrill Tkachov
2019-08-30 14:44 ` Jonathan Wakely
2019-05-15 12:43 ` [ARM/FDPIC v5 09/21] [ARM] FDPIC: Add support for taking address of nested function Christophe Lyon
2019-07-16 11:53 ` Kyrill Tkachov
2019-07-16 13:31 ` Kyrill Tkachov
2019-07-31 14:48 ` Christophe Lyon
2019-08-29 15:40 ` Christophe Lyon
2019-08-30 8:54 ` Kyrill Tkachov [this message]
2019-05-15 12:43 ` [ARM/FDPIC v5 10/21] [ARM] FDPIC: Implement TLS support Christophe Lyon
2019-09-04 14:16 ` Kyrill Tkachov
2019-09-04 20:03 ` Christophe Lyon
2019-09-09 8:54 ` Christophe Lyon
2019-05-15 12:43 ` [ARM/FDPIC v5 08/21] [ARM] FDPIC: Enforce local/global binding for function descriptors Christophe Lyon
2019-07-16 10:51 ` Kyrill Tkachov
2019-05-15 12:44 ` [ARM/FDPIC v5 13/21] [ARM] FDPIC: Force LSB bit for PC in Cortex-M architecture Christophe Lyon
2019-08-29 15:37 ` Kyrill Tkachov
2019-09-05 8:30 ` Christophe Lyon
2019-09-05 8:32 ` Christophe Lyon
2019-09-05 20:56 ` Ian Lance Taylor
2019-09-05 9:03 ` Kyrill Tkachov
2019-09-09 8:58 ` Christophe Lyon
2019-05-15 12:44 ` [ARM/FDPIC v5 11/21] [ARM] FDPIC: Add support to unwind FDPIC signal frame Christophe Lyon
2019-09-04 14:19 ` Kyrill Tkachov
2019-05-15 12:44 ` [ARM/FDPIC v5 12/21] [ARM] FDPIC: Restore r9 after we call __aeabi_read_tp Christophe Lyon
2019-08-29 15:40 ` Christophe Lyon
2019-08-29 15:44 ` Kyrill Tkachov
2019-05-15 12:45 ` [ARM/FDPIC v5 15/21] [ARM][testsuite] FDPIC: Adjust scan-assembler patterns Christophe Lyon
2019-07-19 8:54 ` Kyrill Tkachov
2019-05-15 12:45 ` [ARM/FDPIC v5 16/21] [ARM][testsuite] FDPIC: Skip tests that don't work in PIC mode Christophe Lyon
2019-07-19 8:56 ` Kyrill Tkachov
2019-05-15 12:45 ` [ARM/FDPIC v5 14/21] [ARM][testsuite] FDPIC: Skip unsupported tests Christophe Lyon
2019-07-19 8:52 ` Kyrill Tkachov
2019-05-15 12:46 ` [ARM/FDPIC v5 19/21] [ARM][testsuite] FDPIC: Adjust pr43698.c to avoid clash with uclibc Christophe Lyon
2019-07-19 9:00 ` Kyrill Tkachov
2019-05-15 12:46 ` [ARM/FDPIC v5 18/21] [ARM][testsuite] FDPIC: Enable tests on pie_enabled targets Christophe Lyon
2019-07-19 8:59 ` Kyrill Tkachov
2019-07-22 19:50 ` Mike Stump
2019-05-15 12:46 ` [ARM/FDPIC v5 17/21] [ARM][testsuite] FDPIC: Handle *-*-uclinux* Christophe Lyon
2019-07-19 8:57 ` Kyrill Tkachov
2019-07-22 19:37 ` Mike Stump
2019-05-15 12:47 ` [ARM/FDPIC v5 21/21] [ARM] FDPIC: Handle stack-protector combined patterns Christophe Lyon
2019-07-19 9:40 ` Kyrill Tkachov
2019-05-15 12:47 ` [ARM/FDPIC v5 20/21] [ARM][testsuite] FDPIC: Skip tests using architectures unsupported by FDPIC Christophe Lyon
2019-07-19 9:03 ` Kyrill Tkachov
2019-09-06 8:01 ` Christophe Lyon
2019-09-06 8:28 ` Kyrill Tkachov
2019-09-06 9:10 ` Christophe Lyon
2019-09-06 17:44 ` Christophe Lyon
2019-09-09 8:38 ` Christophe Lyon
2019-05-23 12:46 ` [ARM/FDPIC v5 00/21] FDPIC ABI for ARM Christophe Lyon
2019-06-04 12:57 ` Christophe Lyon
2019-06-06 12:36 ` Christophe Lyon
2019-06-17 11:42 ` Christophe Lyon
2019-07-01 12:16 ` Christophe Lyon
2019-07-08 14:28 ` Christophe Lyon
2019-07-16 9:13 ` Christophe Lyon
2019-08-29 14:54 ` Christophe Lyon
2019-08-29 16:29 ` Christophe Lyon
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=7bad7519-da2b-cf7b-a3be-15c86e6fe27e@foss.arm.com \
--to=kyrylo.tkachov@foss.arm.com \
--cc=christophe.lyon@linaro.org \
--cc=christophe.lyon@st.com \
--cc=gcc-patches@gcc.gnu.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).