From: Simon Marchi <simark@simark.ca>
To: Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>,
gdb-patches@sourceware.org
Subject: Re: [PATCH 06/26] gdbserver: turn part of get_thread_regcache into regcache::fetch
Date: Thu, 21 Dec 2023 15:48:22 -0500 [thread overview]
Message-ID: <5bd0ee41-915f-4bbc-95e7-62baaccbc430@simark.ca> (raw)
In-Reply-To: <d2bd6c9812d5565f7ce91501892680ba54ec4522.1677582744.git.tankut.baris.aktemur@intel.com>
On 2/28/23 06:28, Tankut Baris Aktemur via Gdb-patches wrote:
> Extract out a piece of code in get_thread_regcache that fetches the
> registers from the target, and turn it into a method of 'regcache'.
> ---
> gdbserver/regcache.cc | 23 +++++++++++++++--------
> gdbserver/regcache.h | 3 +++
> 2 files changed, 18 insertions(+), 8 deletions(-)
>
> diff --git a/gdbserver/regcache.cc b/gdbserver/regcache.cc
> index 2a8dc17ed6a..89ecdfec6f3 100644
> --- a/gdbserver/regcache.cc
> +++ b/gdbserver/regcache.cc
> @@ -48,19 +48,26 @@ get_thread_regcache (struct thread_info *thread, bool fetch)
> regcache->thread = thread;
> }
>
> - if (fetch && regcache->registers_valid == 0)
> + if (fetch)
> + regcache->fetch ();
> +
> + return regcache;
> +}
> +
> +void
> +regcache::fetch ()
> +{
> + if (registers_valid == 0)
> {
> scoped_restore_current_thread restore_thread;
> + gdb_assert (this->thread != nullptr);
> + switch_to_thread (this->thread);
>
> - switch_to_thread (thread);
> /* Invalidate all registers, to prevent stale left-overs. */
> - memset (regcache->register_status, REG_UNAVAILABLE,
> - regcache->tdesc->reg_defs.size ());
> - fetch_inferior_registers (regcache, -1);
> - regcache->registers_valid = 1;
> + memset (register_status, REG_UNAVAILABLE, tdesc->reg_defs.size ());
> + fetch_inferior_registers (this, -1);
> + registers_valid = 1;
> }
> -
> - return regcache;
Ok, so this is one of the uses of the regcache::thread field. Given
that, as of today (before your series), a thread knows its regcache, but
a regcache doesn't know its thread (if any), it could be the other way
around. You could have a thread_info::fetch_registers method that does
the above. This way, the regcache stays a relatively dumb box of
register data, oblivious to where the bytes come from. Again, I don't
have yet the full context of the rest of the series.
Another improvement I'd like to see: would it be relatively easy to
change fetch_inferior_registers to pass the thread down, and avoid the
switch_to_thread? I guess you wouldn't need to pass the regcache down
anymore, because the target function would know it needs to put the
register data in thread->regcache. Doing the same in GDB is an enormous
amount of work, but I peeked at the GDBserver code, and it seems
do-able.
Simon
next prev parent reply other threads:[~2023-12-21 20:48 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-02-28 11:27 [PATCH 00/26] gdbserver: refactor regcache and allow gradually populating Tankut Baris Aktemur
2023-02-28 11:27 ` [PATCH 01/26] gdbserver: convert init_register_cache into regcache::initialize Tankut Baris Aktemur
2023-12-21 20:12 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 02/26] gdbserver: convert new_register_cache into a regcache constructor Tankut Baris Aktemur
2023-12-21 20:19 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 03/26] gdbserver: by-pass regcache to access tdesc only Tankut Baris Aktemur
2023-12-21 20:22 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 04/26] gdbserver: boolify and defaultize the 'fetch' parameter of get_thread_regcache Tankut Baris Aktemur
2023-12-21 20:24 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 05/26] gdbserver: add a pointer to the owner thread in regcache Tankut Baris Aktemur
2023-12-21 20:28 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 06/26] gdbserver: turn part of get_thread_regcache into regcache::fetch Tankut Baris Aktemur
2023-12-21 20:48 ` Simon Marchi [this message]
2023-02-28 11:28 ` [PATCH 07/26] gdbserver: convert regcache_cpy into regcache::copy_from Tankut Baris Aktemur
2023-12-21 20:50 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 08/26] gdbserver: convert free_register_cache into a destructor of regcache Tankut Baris Aktemur
2023-12-21 20:57 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 09/26] gdbserver: extract out regcache::invalidate and regcache::discard Tankut Baris Aktemur
2023-12-21 21:08 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 10/26] gdbserver: convert registers_to_string into regcache::registers_to_string Tankut Baris Aktemur
2023-12-21 21:13 ` Simon Marchi
2023-12-21 21:19 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 11/26] gdbserver: convert registers_from_string into regcache::registers_from_string Tankut Baris Aktemur
2023-02-28 11:28 ` [PATCH 12/26] gdbserver: convert supply_regblock to regcache::supply_regblock Tankut Baris Aktemur
2023-12-21 21:23 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 13/26] gdbserver: convert register_data into regcache::register_data Tankut Baris Aktemur
2023-12-21 21:26 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 14/26] gdbserver: introduce and use regcache::set_register_status Tankut Baris Aktemur
2023-12-21 21:30 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 15/26] gdbserver: check for nullptr condition in regcache::get_register_status Tankut Baris Aktemur
2023-12-21 21:32 ` Simon Marchi
2023-12-21 21:34 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 16/26] gdbserver: boolify regcache fields Tankut Baris Aktemur
2023-12-22 3:20 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 17/26] gdbserver: rename regcache's registers_valid to registers_fetched Tankut Baris Aktemur
2023-12-22 3:23 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 18/26] gdbsupport: fix a typo in a comment in common-regcache.h Tankut Baris Aktemur
2023-12-22 3:24 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 19/26] gdbserver: fix the declared type of register_status in regcache Tankut Baris Aktemur
2023-12-22 3:35 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 20/26] gdbserver: make some regcache fields private Tankut Baris Aktemur
2023-12-22 3:39 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 21/26] gdbserver: use REG_UNKNOWN for a regcache's register statuses Tankut Baris Aktemur
2023-12-22 4:32 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 22/26] gdbserver: zero-out register values in regcache-discard Tankut Baris Aktemur
2023-12-22 4:36 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 23/26] gdbserver: set register statuses in registers_from_string Tankut Baris Aktemur
2023-12-22 4:40 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 24/26] gdbserver: return tracked register status in regcache_raw_read_unsigned Tankut Baris Aktemur
2023-12-22 4:42 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 25/26] gdbserver: refuse null argument in regcache::supply_regblock Tankut Baris Aktemur
2023-12-22 4:54 ` Simon Marchi
2023-02-28 11:28 ` [PATCH 26/26] gdbserver: allow gradually populating and selectively storing a regcache Tankut Baris Aktemur
2023-12-22 16:25 ` Simon Marchi
2023-03-07 20:39 ` [PATCH 00/26] gdbserver: refactor regcache and allow gradually populating Tom Tromey
2023-03-13 14:33 ` Aktemur, Tankut Baris
2023-03-28 13:42 ` Aktemur, Tankut Baris
2023-06-20 12:58 ` Aktemur, Tankut Baris
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=5bd0ee41-915f-4bbc-95e7-62baaccbc430@simark.ca \
--to=simark@simark.ca \
--cc=gdb-patches@sourceware.org \
--cc=tankut.baris.aktemur@intel.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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).