From: Simon Marchi <simark@simark.ca>
To: Kirill Radkin <kirill.radkin@syntacore.com>,
"gdb-patches@sourceware.org" <gdb-patches@sourceware.org>
Cc: Konstantin Vladimirov <konstantin.vladimirov@syntacore.com>,
Ivan Tetyushkin <ivan.tetyushkin@syntacore.com>
Subject: Re: [PATCH] [gdbserver] Fix overflow detection in gdbserver
Date: Thu, 21 Dec 2023 14:42:21 -0500 [thread overview]
Message-ID: <477a743c-b318-4089-bfbc-0035686b4cad@simark.ca> (raw)
In-Reply-To: <f4e42985699247b2ab9c2635770063d6@syntacore.com>
On 12/21/23 03:38, Kirill Radkin wrote:
> Hi,
>
>
> Recently I have encountered an issue with gdbserver not being able to send debug information from large files (> 2Gb). I found that this issue is connected with parsing offsets (integer) from vFile::pread packets. There is already a patch (https://sourceware.org/bugzilla/show_bug.cgi?id=23198) that allows to parse integers up to 0x7fffffff (32-bit integer), but it doesn't fix the problem at all, because offsets can have a off_t type which can be larger than 32-bit integer (it depends on system).
>
> This patch allows require_int() function to parse offset up to the maximum value implied by the off_t type.
Hi,
Thanks for the patch.
First, the boring stuff: I think that the patch is more than an obvious
fix, so we'll need to make sure you are covered with an FSF copyright
assignment [1]. Do you know if you (or your employer, since it looks
like you're doing this as part of your employment) are covered with one
already?
[1] https://sourceware.org/gdb/wiki/ContributionChecklist#FSF_copyright_Assignment
Then, I am unable to apply your patch with git-am. It looks like you
copy pasted the patch in your email client, which usually gives bad
results. Are you able to use git-send-email [2]? This is the most
trustworthy way.
[2] https://sourceware.org/gdb/wiki/ContributionChecklist#Submitting_patches
Some comments below.
> From 01117d757a0eb4f632aded4b15a06dbcccc7adf7 Mon Sep 17 00:00:00 2001
> From: Kirill Radkin <kirill.radkin@syntacore.com>
> Date: Mon, 13 Nov 2023 16:27:15 +0300
> Subject: [PATCH] gdbserver: Fix overflow detection in gdbserver
>
> Currently gdbserver use require_int() function to
> parse offset which we get, for example, in vFile::pread
> packet. This function allows integers up to 0x7fffffff,
> (to fit in 32-bit int) but actually offset (for pread
> system call) have a off_t type which can be larger than
> 32-bit.
>
> This patch allows require_int() function to parse offset
> up to the maximum value implied by the off_t type.
> ---
> gdb/testsuite/gdb.server/pread-offset-size.S | 26 ++++++++++
> .../gdb.server/pread-offset-size.exp | 47 +++++++++++++++++++
> gdbserver/hostio.cc | 16 +++++--
> 3 files changed, 85 insertions(+), 4 deletions(-)
> create mode 100644 gdb/testsuite/gdb.server/pread-offset-size.S
> create mode 100644 gdb/testsuite/gdb.server/pread-offset-size.exp
>
> diff --git a/gdb/testsuite/gdb.server/pread-offset-size.S b/gdb/testsuite/gdb.server/pread-offset-size.S
> new file mode 100644
> index 00000000000..31748090ac3
> --- /dev/null
> +++ b/gdb/testsuite/gdb.server/pread-offset-size.S
> @@ -0,0 +1,26 @@
> +/* This testcase is part of GDB, the GNU debugger.
> +
> + Copyright 2023-2023 Free Software Foundation, Inc.
Just "2023".
> +
> + This program is free software; you can redistribute it and/or modify
> + it under the terms of the GNU General Public License as published by
> + the Free Software Foundation; either version 3 of the License, or
> + (at your option) any later version.
> +
> + This program is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + GNU General Public License for more details.
> +
> + You should have received a copy of the GNU General Public License
> + along with this program. If not, see <http://www.gnu.org/licenses/>. */
> +
> +.text
> +.globl_start
> +_start:
> +.skip 3742415472
Does this create a sparse file on systems that support it? I could see
it as a problem for some test environments if we create and leave a file
occupying ~4GB file in the build directory. Well, two actually, there
will be the .o and the final executable. Note that I don't have a
better idea on how to test this...
> + ret
> +.globlf
> +.typef, @function
> +f:
> + ret
Do you expect this .S file to work for all architectures? It's simple,
but I guess that the "ret" instruction doesn't exist on all
architectures. Can you think of a way to produce a big binary that is
not architecture-dependent at all? Maybe produce an executable, and
then insert a large dummy section in the beginning?
> diff --git a/gdb/testsuite/gdb.server/pread-offset-size.exp b/gdb/testsuite/gdb.server/pread-offset-size.exp
> new file mode 100644
> index 00000000000..a4b648b29fc
> --- /dev/null
> +++ b/gdb/testsuite/gdb.server/pread-offset-size.exp
> @@ -0,0 +1,47 @@
> +# Copyright (C) 2023-2023 Free Software Foundation, Inc.
Just "2023".
> +#
> +# This program is free software; you can redistribute it and/or modify
> +# it under the terms of the GNU General Public License as published by
> +# the Free Software Foundation; either version 3 of the License, or
> +# (at your option) any later version.
> +#
> +# This program is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> +# GNU General Public License for more details.
> +#
> +# You should have received a copy of the GNU General Public License
> +# along with this program. If not, see <http://www.gnu.org/licenses/>.
> +
Please add a short comment that explains the intention of this test.
> +load_lib gdbserver-support.exp
> +
> +if {[skip_gdbserver_tests]} {
> + return
> +}
> +
> +standard_testfile .S
> +
> +if { [prepare_for_testing ${testfile}.exp $testfile \
> + $srcfile {debug additional_flags=-nostdlib} ] } {
> + return -1
> +}
> +
> +gdb_exit
> +gdb_start
> +
> +gdb_test_no_output "set remote exec-file $binfile" \
> +"set remote exec-file"
> +
> +# Make sure we're disconnected, in case we're testing with an
> +# extended-remote board, therefore already connected.
> +gdb_test "disconnect" ".*"
> +
> +set res [gdbserver_spawn ""]
> +set gdbserver_protocol [lindex $res 0]
> +set gdbserver_gdbport [lindex $res 1]
> +
> +gdb_test "target $gdbserver_protocol $gdbserver_gdbport" \
> +"Remote debugging using .*" \
> +"target $gdbserver_protocol $gdbserver_gdbport"
> +
> +gdb_test "break f" "Breakpoint 1.*"
> diff --git a/gdbserver/hostio.cc b/gdbserver/hostio.cc
> index ea70c26da0f..068771428f9 100644
> --- a/gdbserver/hostio.cc
> +++ b/gdbserver/hostio.cc
> @@ -90,12 +90,16 @@ require_filename (char **pp, char *filename)
> return 0;
> }
>
> +template <typename T>
> static int
> -require_int (char **pp, int *value)
> +require_int (char **pp, T *value)
> {
> char *p;
> int count, firstdigit;
>
> + /* Max count of hexadecimal digits in off_t (1 hex digit is 4 bits) */
> + int max_count = sizeof(T) * CHAR_BIT / 4;
> +
> p = *pp;
> *value = 0;
> count = 0;
> @@ -112,7 +116,9 @@ require_int (char **pp, int *value)
> firstdigit = nib;
>
> /* Don't allow overflow. */
> - if (count >= 8 || (count == 7 && firstdigit >= 0x8))
> + if (count >= max_count || (static_cast<T>(-1) < 0
> +&& count == (max_count - 1)
> +&& firstdigit >= 0x8))
> return -1;
I guess that the static_cast thing is to determine if the type is signed
or not? Can you give it a name, for clarity, like (maybe there's a
better way, just an example):
constexpr bool is_signed = static_cast<T> (-1) < 0;
Simon
next prev parent reply other threads:[~2023-12-21 19:42 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-21 8:38 Kirill Radkin
2023-12-21 19:42 ` Simon Marchi [this message]
2023-12-22 15:41 ` Tom Tromey
2024-01-10 14:57 ` [PATCH] gdbserver: " Kirill Radkin
2024-02-08 16:19 ` Tom Tromey
2024-02-19 14:19 ` [PATCH v2] " Kirill Radkin
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=477a743c-b318-4089-bfbc-0035686b4cad@simark.ca \
--to=simark@simark.ca \
--cc=gdb-patches@sourceware.org \
--cc=ivan.tetyushkin@syntacore.com \
--cc=kirill.radkin@syntacore.com \
--cc=konstantin.vladimirov@syntacore.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).