This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 6/9] Support for DTrace USDT probes in x86_64 targets.
- From: Sergio Durigan Junior <sergiodj at redhat dot com>
- To: "Jose E. Marchesi" <jose dot marchesi at oracle dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Wed, 08 Oct 2014 15:32:51 -0400
- Subject: Re: [PATCH 6/9] Support for DTrace USDT probes in x86_64 targets.
- Authentication-results: sourceware.org; auth=none
- References: <1411724905-31234-1-git-send-email-jose dot marchesi at oracle dot com> <1411724905-31234-7-git-send-email-jose dot marchesi at oracle dot com>
On Friday, September 26 2014, Jose E. Marchesi wrote:
> This patch adds the target-specific code in order to support the
> calculation of DTrace probes arguments in x86_64 targets, and also the
> enabling and disabling of probes. This is done by implementing the
> `dtrace_*' gdbarch handlers.
Thanks. Comments below.
> gdb:
>
> 2014-09-26 Jose E. Marchesi <jose.marchesi@oracle.com>
>
> * amd64-linux-tdep.h: Prototypes for
> `amd64_dtrace_probe_argument', `amd64_dtrace_enable_probe',
> `amd64_dtrace_disable_probe' and `gdbarch_dtrace_probe_is_enabled'.
>
> * amd64-linux-tdep.c (amd64_dtrace_probe_argument): New function.
> (amd64_dtrace_probe_is_enabled): Likewise.
> (amd64_dtrace_enable_probe): Likewise.
> (amd64_dtrace_disable_probe): Likewise.
> (amd64_linux_init_abi): Register the
> `gdbarch_dtrace_probe_argument', `gdbarch_dtrace_enable_probe',
> `gdbarch_dtrace_disable_probe' and `gdbarch_dtrace_probe_is_enabled'.
> ---
> gdb/ChangeLog | 14 +++++
> gdb/amd64-linux-tdep.c | 150 ++++++++++++++++++++++++++++++++++++++++++++++++
> gdb/amd64-linux-tdep.h | 11 ++++
> 3 files changed, 175 insertions(+)
>
> diff --git a/gdb/ChangeLog b/gdb/ChangeLog
> index eac03e7..a32d303 100644
> --- a/gdb/ChangeLog
> +++ b/gdb/ChangeLog
> @@ -1,5 +1,19 @@
> 2014-09-26 Jose E. Marchesi <jose.marchesi@oracle.com>
>
> + * amd64-linux-tdep.h: Prototypes for
> + `amd64_dtrace_probe_argument', `amd64_dtrace_enable_probe',
> + `amd64_dtrace_disable_probe' and `gdbarch_dtrace_probe_is_enabled'.
> +
> + * amd64-linux-tdep.c (amd64_dtrace_probe_argument): New function.
> + (amd64_dtrace_probe_is_enabled): Likewise.
> + (amd64_dtrace_enable_probe): Likewise.
> + (amd64_dtrace_disable_probe): Likewise.
> + (amd64_linux_init_abi): Register the
> + `gdbarch_dtrace_probe_argument', `gdbarch_dtrace_enable_probe',
> + `gdbarch_dtrace_disable_probe' and `gdbarch_dtrace_probe_is_enabled'.
> +
> +2014-09-26 Jose E. Marchesi <jose.marchesi@oracle.com>
> +
> * breakpoint.c (BREAK_ARGS_HELP): help string updated to mention
> the -probe-dtrace new vpossible value for PROBE_MODIFIER.
>
> diff --git a/gdb/amd64-linux-tdep.c b/gdb/amd64-linux-tdep.c
> index 850ca20..273f5c4 100644
> --- a/gdb/amd64-linux-tdep.c
> +++ b/gdb/amd64-linux-tdep.c
> @@ -28,6 +28,8 @@
> #include "gdbtypes.h"
> #include "reggroups.h"
> #include "regset.h"
> +#include "parser-defs.h"
> +#include "user-regs.h"
> #include "amd64-linux-tdep.h"
> #include "i386-linux-tdep.h"
> #include "linux-tdep.h"
> @@ -1609,6 +1611,148 @@ amd64_linux_core_read_description (struct gdbarch *gdbarch,
> }
> }
>
> +/* Implementation of `gdbarch_dtrace_probe_is_enabled', as defined in
> + gdbarch.h. */
> +
> +int
> +amd64_dtrace_probe_is_enabled (struct gdbarch *gdbarch, CORE_ADDR addr)
> +{
> + /* The instruction sequence used in x86_64 machines for a disabled
> + is-enabled probe is:
> +
> + xor %rax, %rax => 48 33 C0
> + ADDR: nop => 90
> + nop => 90
> +
> + or
> +
> + xor %rax, %rax => 48 33 C0
> + ADDR: ret => c3
> + nop => 90
> +
> + This function returns 1 if the instructions at ADDR do _not_
> + follow any of these patterns.
> +
> + Note that ADDR is offset 3 bytes from the beginning of these
> + sequences. */
This comment could be placed on the top of the function (after the
"Implementation of ..."), or after the declaration of the variables.
Sorry for the nitpick, but I find it strange to make the comment before
declaring the variables in the body of the function (and I don't
remember seeing this on GDB very frequently).
> + gdb_byte buf[5];
> + read_memory (addr - 3, buf, 5);
> +
> + return !((buf[0] == 0x48) && (buf[1] == 0x33) && (buf[2] == 0xc0) /* xor */
> + && ((buf[3] == 0x90) || (buf[3] == 0xc3)) /* nop | ret */
> + && (buf[4] == 0x90)); /* nop */
> +}
> +
> +/* Implementation of `gdbarch_dtrace_enable_probe', as defined in
> + gdbarch.h. */
> +
> +void
> +amd64_dtrace_enable_probe (struct gdbarch *gdbarch, CORE_ADDR addr)
> +{
> + /* We use the following instruction sequence for enabling an
> + is-enabled probe:
> +
> + mov $0x1, %eax => b8 01 00 00 00
> +
> + Note also that ADDR is offset 3 bytes from the beginning of the
> + sequence. */
> +
> + gdb_byte buf[5];
> +
> + buf[0] = 0xb8; buf[1] = 0x01; buf[2] = 0x00; buf[3] = 0x00; buf[4] = 0x00;
Each assignment should go on its own line.
> + write_memory (addr - 3, buf, 5);
> +}
> +
> +/* Implementation of `gdbarch_dtrace_disable_probe', as defined in
> + gdbarch.h. */
> +
> +void
> +amd64_dtrace_disable_probe (struct gdbarch *gdbarch, CORE_ADDR addr)
> +{
> + /* We use the following instruction sequence for disabling an
> + is-enabled probe:
> +
> + xor %rax, %rax; nop; nop => 48 33 C0 90 90
> +
> + Note that ADDR is offset 3 bytes from the beginning of the
> + sequence. */
> +
> + gdb_byte buf[5];
> +
> + buf[0] = 0x48; buf[1] = 0x33; buf[2] = 0xc0; buf[3] = 0x90; buf[4] = 0x90;
Likewise.
> + write_memory (addr - 3, buf, 5);
> +}
> +
> +/* Implementation of `gdbarch_dtrace_parse_special_token', as defined in
> + gdbarch.h. */
This comment needs an update, there is no
"gdbarch_dtrace_parse_special_token" :-).
> +
> +void
> +amd64_dtrace_probe_argument (struct gdbarch *gdbarch,
> + struct parser_state *pstate,
> + int narg)
> +{
> + static int arg_reg_map[6] =
> + {
> + AMD64_RDI_REGNUM, /* Arg 1. */
> + AMD64_RSI_REGNUM, /* Arg 2. */
> + AMD64_RDX_REGNUM, /* Arg 3. */
> + AMD64_RCX_REGNUM, /* Arg 4. */
> + AMD64_R8_REGNUM, /* Arg 5. */
> + AMD64_R9_REGNUM /* Arg 6. */
> + };
> +
> + enum bfd_endian byte_order = gdbarch_byte_order (gdbarch);
> + struct frame_info *this_frame = get_selected_frame (NULL);
> + struct stoken str;
No newline between variables being declared.
> +
> + /* DTrace probe arguments can be found on the ABI-defined places for
> + regular arguments at the current PC. The probe abstraction
> + currently supports up to 12 arguments for probes. */
> +
> + if (narg < 6)
> + {
> + int regno = arg_reg_map [narg];
No space between variable name and index operator "[".
> + const char *regname = user_reg_map_regnum_to_name (gdbarch, regno);
> +
> + write_exp_elt_opcode (pstate, OP_REGISTER);
> + str.ptr = regname;
> + str.length = strlen (regname);
> + write_exp_string (pstate, str);
> + write_exp_elt_opcode (pstate, OP_REGISTER);
> + }
> + else
> + {
> + /* Additional arguments are passed on the stack. */
> +
> + CORE_ADDR sp;
Spurious newline.
> + const char *regname = user_reg_map_regnum_to_name (gdbarch, AMD64_RSP_REGNUM);
> +
> + /* Displacement. */
> + write_exp_elt_opcode (pstate, OP_LONG);
> + write_exp_elt_type (pstate, builtin_type (gdbarch)->builtin_long);
> + write_exp_elt_longcst (pstate, narg - 6);
> + write_exp_elt_opcode (pstate, OP_LONG);
As I said in another message, I'm not a fan of aligning the open paren
for function calls; it messes with a simple grep, for example.
> +
> + /* Register: SP. */
> + write_exp_elt_opcode (pstate, OP_REGISTER);
> + str.ptr = regname;
> + str.length = strlen (regname);
> + write_exp_string (pstate, str);
> + write_exp_elt_opcode (pstate, OP_REGISTER);
> +
> + write_exp_elt_opcode (pstate, BINOP_ADD);
> +
> + /* Cast to long. */
> + write_exp_elt_opcode (pstate, UNOP_CAST);
> + write_exp_elt_type (pstate,
> + lookup_pointer_type (builtin_type (gdbarch)->builtin_long));
> + write_exp_elt_opcode (pstate, UNOP_CAST);
> +
> + write_exp_elt_opcode (pstate, UNOP_IND);
> + }
> +}
> +
> static void
> amd64_linux_init_abi_common(struct gdbarch_info info, struct gdbarch *gdbarch)
> {
> @@ -1872,6 +2016,12 @@ amd64_linux_init_abi (struct gdbarch_info info, struct gdbarch *gdbarch)
> /* GNU/Linux uses SVR4-style shared libraries. */
> set_solib_svr4_fetch_link_map_offsets
> (gdbarch, svr4_lp64_fetch_link_map_offsets);
> +
> + /* Register DTrace handlers. */
> + set_gdbarch_dtrace_probe_argument (gdbarch, amd64_dtrace_probe_argument);
> + set_gdbarch_dtrace_probe_is_enabled (gdbarch, amd64_dtrace_probe_is_enabled);
> + set_gdbarch_dtrace_enable_probe (gdbarch, amd64_dtrace_enable_probe);
> + set_gdbarch_dtrace_disable_probe (gdbarch, amd64_dtrace_disable_probe);
> }
>
> static void
> diff --git a/gdb/amd64-linux-tdep.h b/gdb/amd64-linux-tdep.h
> index 25563b8..b28dc50 100644
> --- a/gdb/amd64-linux-tdep.h
> +++ b/gdb/amd64-linux-tdep.h
> @@ -594,4 +594,15 @@ enum amd64_x32_syscall {
> amd64_x32_sys_getsockopt = (amd64_x32_syscall_bit + 542),
> };
>
> +/* DTrace related functions. */
> +
> +extern void amd64_dtrace_probe_argument (struct gdbarch *gdbarch,
> + struct parser_state *pstate,
> + int narg);
> +
> +extern int amd64_dtrace_probe_is_enabled (struct gdbarch *gdbarch, CORE_ADDR addr);
> +
> +extern void amd64_dtrace_enable_probe (struct gdbarch *gdbarch, CORE_ADDR addr);
> +extern void amd64_dtrace_disable_probe (struct gdbarch *gdbarch, CORE_ADDR addr);
> +
> #endif /* amd64-linux-tdep.h */
> --
> 1.7.10.4
Sorry for not being able to do an extensive review, but dtrace is
proprietary and I know very little about its architecture.
--
Sergio
GPG key ID: 0x65FC5E36
Please send encrypted e-mail if possible
http://sergiodj.net/