[11/16] gdb/gdbserver: pass inferior arguments as a single string
Commit Message
GDB holds the inferior arguments as a single string. Currently when
GDB needs to pass the inferior arguments to a remote target as part of
a vRun packet, this is done by splitting the single argument string
into its component arguments by calling gdb::remote_args::split, which
uses the gdb_argv class to split the arguments for us.
The same gdb_argv class is used when the user has asked GDB/gdbserver
to start the inferior without first invoking a shell; the gdb_argv
class is used to split the argument string into it component
arguments, and each is passed as a separate argument to the execve
call which spawns the inferior.
There is however, a problem with using gdb_argv to split the arguments
before passing them to a remote target. To understand this problem we
must first understand how gdb_argv is used when invoking an inferior
without a shell.
And to understand how gdb_argv is used to start an inferior without a
shell, I feel we need to first look at an example of starting an
inferior with a shell.
Consider these two cases:
(a) (gdb) set args \$VAR
(b) (gdb) set args $VAR
When starting with a shell, in case (a) the user expects the inferior
to receive a literal '$VAR' string as an argument, while in case (b)
the user expects to see the shell expanded value of the variable $VAR.
If the user does 'set startup-with-shell off', then in (a) GDB will
strip the '\' while splitting the arguments, and the inferior will be
passed a literal '$VAR'. In (b) there is no '\' to strip, so also in
this case the inferior will receive a literal '$VAR', remember
startup-with-shell is off, so there is no shell than can ever expand
$VAR.
Notice, that when startup-with-shell is off, we end up with a many to
one mapping, both (a) and (b) result in the literal string $VAR being
passed to the inferior. I think this is the correct behaviour in this
case.
However, as we use gdb_argv to split the remote arguments we have the
same many to one mapping within the vRun packet. But the vRun packet
will be used when startup-with-shell is both on and off. What this
means is that when gdbserver receives a vRun packet containing '$VAR'
it doesn't know if GDB actually had '$VAR', or if GDB had '\$VAR'.
And this is a huge problem.
We can try to address this by making the argument splitting for remote
targets smarter. And later in this series I will do that. However, I
think that splitting and joining the arguments as we do was a
mistake. The later patch in this series handles unquoted, single
quoted, and double quoted strings. But doesn't really address
parameter substitution, command substitution, or arithmetic
expansion. And even if we did try to address these cases, what rules
exactly would we implement? Probably POSIX shell rules, but what if
the remote target doesn't have a POSIX shell? Why do we need to pick
an particular rule set?
Clearly, for backward compatibility we need to maintain some degree of
argument splitting and joining as we currently have; and that's why I
have a later patch in this series that tries to improve that splitting
and joining a little. But I think, what we should really do, is add a
new feature flag (as used by the qSupported packet) and, if GDB and
the remote target agree, we should pass the inferior arguments as a
single string.
This solves all our problems. In the startup with shell case, we no
longer need to worry about splitting at all. The arguments are passed
unmodified to the remote target, who can then pass the arguments to
the shell directly.
In the 'startup-with-shell off' case it is now up to the remote target
to split the arguments, though in gdbserver we already did this as we
always joined the arguments, so for gdbserver it's no significant
hardship.
And if the remote target doesn't have a POSIX shell, well GDB just
doesn't need to worry about it!
Something similar to this was originally suggested in this series:
https://inbox.sourceware.org/gdb-patches/20211022071933.3478427-1-m.weghorn@posteo.de/
though this series didn't try to maintain backward compatibility,
which I think is an issue that my patch solves. Additionally, this
series only passed the arguments as a single string in some cases,
I've simplified this so that when GDB and the remote agree, the
arguments are always passed as a single string. I think this makes my
version a little cleaner.
I've also added documentation and some tests with this commit,
including ensuring that we test both the new single string approach,
and the fallback split/join approach.
I've credited the author of the referenced series as co-author as they
did come to a similar conclusion, though I think my implementation is
different enough that I'm happy to list myself as primary author.
Co-Authored-By: Michael Weghorn <m.weghorn@posteo.de>
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=28392
---
gdb/NEWS | 7 +
gdb/doc/gdb.texinfo | 23 +++
gdb/remote.c | 25 ++-
gdb/testsuite/gdb.base/args.exp | 47 +++--
gdb/testsuite/gdb.base/inferior-args.exp | 31 +++-
gdb/testsuite/gdb.base/startup-with-shell.exp | 161 ++++++++++--------
gdbserver/server.cc | 21 ++-
gdbserver/server.h | 5 +
8 files changed, 226 insertions(+), 94 deletions(-)
Comments
> From: Andrew Burgess <aburgess@redhat.com>
> Cc: Andrew Burgess <aburgess@redhat.com>, Michael Weghorn <m.weghorn@posteo.de>
> Date: Tue, 9 Jan 2024 14:26:34 +0000
>
> gdb/NEWS | 7 +
> gdb/doc/gdb.texinfo | 23 +++
> gdb/remote.c | 25 ++-
> gdb/testsuite/gdb.base/args.exp | 47 +++--
> gdb/testsuite/gdb.base/inferior-args.exp | 31 +++-
> gdb/testsuite/gdb.base/startup-with-shell.exp | 161 ++++++++++--------
> gdbserver/server.cc | 21 ++-
> gdbserver/server.h | 5 +
> 8 files changed, 226 insertions(+), 94 deletions(-)
The documentation parts are okay, thanks.
> From: Andrew Burgess <aburgess@redhat.com>
> Cc: Andrew Burgess <aburgess@redhat.com>, Michael Weghorn <m.weghorn@posteo.de>
> Date: Tue, 9 Jan 2024 14:26:34 +0000
>
> gdb/NEWS | 7 +
> gdb/doc/gdb.texinfo | 23 +++
> gdb/remote.c | 25 ++-
> gdb/testsuite/gdb.base/args.exp | 47 +++--
> gdb/testsuite/gdb.base/inferior-args.exp | 31 +++-
> gdb/testsuite/gdb.base/startup-with-shell.exp | 161 ++++++++++--------
> gdbserver/server.cc | 21 ++-
> gdbserver/server.h | 5 +
> 8 files changed, 226 insertions(+), 94 deletions(-)
The documentation parts are okay, thanks.
Reviewed-By: Eli Zaretskii <eliz@gnu.org>
@@ -120,6 +120,13 @@ QThreadOptions in qSupported
QThreadOptions packet, and the qSupported response can contain the
set of thread options the remote stub supports.
+single-inf-arg in qSupported
+ The new single-inf-arg feature within the qSupported packet allows
+ GDB to inform the stub that it would like to send the inferior
+ arguments as a single string within the vRun packet. The stub can
+ reply with the single-inf-arg feature to indicate that it is able to
+ accept arguments as a single string.
+
*** Changes in GDB 14
* GDB now supports the AArch64 Scalable Matrix Extension 2 (SME2), which
@@ -43041,6 +43041,12 @@
(e.g.@: the last program run). The program is created in the stopped
state.
+If @value{GDBN} sent the @samp{single-inf-arg+} feature in the
+@samp{qSupported} packet, and the stub replied with
+@samp{single-inf-arg+}, then there will only be a single
+@var{argument} string, which includes all inferior arguments,
+separated with whitespace.
+
@c FIXME: What about non-stop mode?
This packet is only available in extended mode (@pxref{extended mode}).
@@ -44471,6 +44477,13 @@
@item vContSupported
This feature indicates whether @value{GDBN} wants to know the
supported actions in the reply to @samp{vCont?} packet.
+
+@item single-inf-arg
+This feature indicates that @value{GDBN} would like to send the
+inferior arguments as a single string within the @samp{vRun} packet.
+@value{GDBN} will not send the arguments as a single string unless the
+stub also reports that is supports this behaviour by including
+@samp{single-inf-arg+} in its @samp{qSupported} reply.
@end table
Stubs should ignore any unknown values for
@@ -44754,6 +44767,11 @@
@tab @samp{-}
@tab No
+@item @samp{single-inf-arg}
+@tab No
+@tab @samp{-}
+@tab No
+
@end multitable
These are the currently defined stub features, in more detail:
@@ -44985,6 +45003,11 @@
@file{/proc/@var{pid}/smaps} file so memory mapping page flags can be inspected.
This is done via the @samp{vFile} requests.
+@item single-inf-arg
+The remote stub would like to receive the inferior arguments as a
+single string within the @samp{vRun} packet. The stub should only
+send this feature if @value{GDBN} sent @samp{single-inf-arg+} in the
+@samp{qSupported} packet.
@end table
@item qSymbol::
@@ -306,6 +306,10 @@ enum {
packets and the tag violation stop replies. */
PACKET_memory_tagging_feature,
+ /* Not really a packet; this indicates support for sending the vRun
+ inferior arguments as a single string. */
+ PACKET_vRun_single_argument,
+
PACKET_MAX
};
@@ -727,6 +731,11 @@ struct remote_features
bool remote_memory_tagging_p () const
{ return packet_support (PACKET_memory_tagging_feature) == PACKET_ENABLE; }
+ /* Returns true if there is support for sending vRun inferior arguments
+ as a single string. */
+ bool remote_vrun_single_arg_p () const
+ { return packet_support (PACKET_vRun_single_argument) == PACKET_ENABLE; }
+
/* Reset all packets back to "unknown support". Called when opening a
new connection to a remote target. */
void reset_all_packet_configs_support ();
@@ -5724,6 +5733,8 @@ static const struct protocol_feature remote_protocol_features[] = {
{ "no-resumed", PACKET_DISABLE, remote_supported_packet, PACKET_no_resumed },
{ "memory-tagging", PACKET_DISABLE, remote_supported_packet,
PACKET_memory_tagging_feature },
+ { "single-inf-arg", PACKET_DISABLE, remote_supported_packet,
+ PACKET_vRun_single_argument },
};
static char *remote_support_xml;
@@ -5835,6 +5846,10 @@ remote_target::remote_query_supported ()
!= AUTO_BOOLEAN_FALSE)
remote_query_supported_append (&q, "memory-tagging+");
+ if (m_features.packet_set_cmd_state (PACKET_vRun_single_argument)
+ != AUTO_BOOLEAN_FALSE)
+ remote_query_supported_append (&q, "single-inf-arg+");
+
/* Keep this one last to work around a gdbserver <= 7.10 bug in
the qSupported:xmlRegisters=i386 handling. */
if (remote_support_xml != NULL
@@ -10626,7 +10641,11 @@ remote_target::extended_remote_run (const std::string &args)
if (!args.empty ())
{
- std::vector<std::string> split_args = gdb::remote_args::split (args);
+ std::vector<std::string> split_args;
+ if (!m_features.remote_vrun_single_arg_p ())
+ split_args = gdb::remote_args::split (args);
+ else
+ split_args.push_back (args);
for (const auto &a : split_args)
{
@@ -15990,6 +16009,10 @@ Show the maximum size of the address (in bits) in a memory packet."), NULL,
add_packet_config_cmd (PACKET_memory_tagging_feature,
"memory-tagging-feature", "memory-tagging-feature", 0);
+ add_packet_config_cmd (PACKET_vRun_single_argument,
+ "single-inferior-argument-feature",
+ "single-inferior-argument-feature", 0);
+
/* Assert that we've registered "set remote foo-packet" commands
for all packet configs. */
{
@@ -72,31 +72,48 @@ proc args_test { name arglist {re_list {}} } {
}
}
-# Test that the --args are processed correctly.
+# Run all the tests.
+proc run_all_tests {} {
+ # Test that the --args are processed correctly.
-args_test basic {{1} {3}}
+ args_test basic {{1} {3}}
-# Test that the --args are processed correctly even if one of them is
-# empty.
+ # Test that the --args are processed correctly even if one of them is
+ # empty.
-args_test "one empty" {{1} {} {3}}
+ args_test "one empty" {{1} {} {3}}
-# Try with 2 empty args.
+ # Try with 2 empty args.
-args_test "two empty" {{1} {} {} 3}
+ args_test "two empty" {{1} {} {} 3}
-# Try with arguments containing literal single quotes.
+ # Try with arguments containing literal single quotes.
-args_test "one empty with single quotes" {{1} {''} {3}}
+ args_test "one empty with single quotes" {{1} {''} {3}}
-args_test "two empty with single quotes" {{1} {''} {''} {3}}
+ args_test "two empty with single quotes" {{1} {''} {''} {3}}
-# Try with arguments containing literal newlines.
+ # Try with arguments containing literal newlines.
-args_test "one newline" {{1} "\n" {3}} {1 \\\\n 3}
+ args_test "one newline" {{1} "\n" {3}} {1 \\\\n 3}
-args_test "two newlines" {{1} "\n" "\n" {3}} {1 \\\\n \\\\n 3}
+ args_test "two newlines" {{1} "\n" "\n" {3}} {1 \\\\n \\\\n 3}
-args_test "lone single quote" {{1} \' {3}}
+ args_test "lone single quote" {{1} \' {3}}
-args_test "lone double quote" {{1} \" {3}} {1 \\\\\" 3}
+ args_test "lone double quote" {{1} \" {3}} {1 \\\\\" 3}
+}
+
+run_all_tests
+
+# For extended-remote targets, disable the packet which passes
+# inferior arguments as a single string. This changes how the vRun
+# (extended-remote only) packet works.
+if {[target_info gdb_protocol] == "extended-remote"} {
+ with_test_prefix "single-inferior-arg disabled" {
+ save_vars { GDBFLAGS } {
+ append GDBFLAGS " -ex \"set remote single-inferior-argument-feature-packet off\""
+ run_all_tests
+ }
+ }
+}
@@ -211,14 +211,31 @@ lappend test_desc_list [list "test four" \
[list "$hex \"'\"" \
"$hex \"\\\\\"\""]]
-foreach desc $test_desc_list {
- lassign $desc name stub_suitable args re_list
- with_test_prefix $name {
- foreach_with_prefix set_method { "start" "starti" "run" "set args" } {
- foreach_with_prefix startup_with_shell { on off } {
- do_test $set_method $startup_with_shell $args $re_list \
- $stub_suitable
+# Run all tests in the global TEST_DESC_LIST.
+proc run_all_tests {} {
+ foreach desc $::test_desc_list {
+ lassign $desc name stub_suitable args re_list
+ with_test_prefix $name {
+ foreach_with_prefix set_method { "start" "starti" "run" "set args" } {
+ foreach_with_prefix startup_with_shell { on off } {
+ do_test $set_method $startup_with_shell $args $re_list \
+ $stub_suitable
+ }
}
}
}
}
+
+run_all_tests
+
+# For extended-remote targets, disable the packet which passes
+# inferior arguments as a single string. This changes how the vRun
+# (extended-remote only) packet works.
+if {[target_info gdb_protocol] == "extended-remote"} {
+ with_test_prefix "single-inferior-arg disabled" {
+ save_vars { GDBFLAGS } {
+ append GDBFLAGS " -ex \"set remote single-inferior-argument-feature-packet off\""
+ run_all_tests
+ }
+ }
+}
@@ -90,76 +90,97 @@ proc run_test_same { args re testname } {
run_test $args $re $re $testname
}
-# The regexp to match a single '\' character.
-set bs "\\\\"
-
-# Are we using 'remote' or 'extended-remote' protocol?
-set is_remote_p [expr [string equal [target_info gdb_protocol] \
- "remote"] \
- || [string equal [target_info gdb_protocol] \
- "extended-remote"]]
-
-## Run the actual tests
-
-run_test "$unique_file_dir/*.unique-extension" \
- "\"$unique_file\"" \
- "\"$unique_file_dir/\\\*\.unique-extension\"" \
- "arg is glob" \
- $is_remote_p
-
-run_test_same "$unique_file_dir/\\*.unique-extension" \
- "\"$unique_file_dir/\\\*\.unique-extension\"" \
- "arg is escaped glob"
-
-save_vars { env(TEST) } {
- set env(TEST) "1234"
- run_test "\$TEST" \
- "\"1234\"" \
- "\"\\\$TEST\"" \
- "arg is shell variable" \
- $is_remote_p
-
- run_test_same "\\\$TEST" \
- "\"\\\$TEST\"" \
- "arg is escaped shell variable"
-}
-
-run_test_same "\"\\a\"" \
- "\"${bs}${bs}a\"" \
- "retain backslash in double quote arg"
-
-run_test_same "'\\a'" \
- "\"${bs}${bs}a\"" \
- "retain backslash in single quote arg"
-
-run_test_same "\"\\\$\"" \
- "\"\\\$\"" \
- "'\$' can be escaped in double quote arg"
-
-run_test_same "'\\\$'" \
- "\"${bs}${bs}\\\$\"" \
- "'\$' is not escaped in single quote arg"
-
-run_test_same "\"\\`\"" \
- "\"\\`\"" \
- "'`' can be escaped in double quote arg"
-
-run_test_same "'\\`'" \
- "\"${bs}${bs}`\"" \
- "'`' is not escaped in single quote arg"
-
-run_test_same "\"\\\"\"" \
- "\"${bs}\"\"" \
- "'\"' can be escaped in double quote arg"
+# Run the actual tests
+proc run_all_tests { { is_remote_with_split_args false } } {
+ # The regexp to match a single '\' character.
+ set bs "\\\\"
+
+ run_test "$::unique_file_dir/*.unique-extension" \
+ "\"$::unique_file\"" \
+ "\"$::unique_file_dir/\\\*\.unique-extension\"" \
+ "arg is glob" \
+ $is_remote_with_split_args
+
+ run_test_same "$::unique_file_dir/\\*.unique-extension" \
+ "\"$::unique_file_dir/\\\*\.unique-extension\"" \
+ "arg is escaped glob"
+
+ save_vars { ::env(TEST) } {
+ set ::env(TEST) "1234"
+ run_test "\$TEST" \
+ "\"1234\"" \
+ "\"\\\$TEST\"" \
+ "arg is shell variable" \
+ $is_remote_with_split_args
+
+ run_test_same "\\\$TEST" \
+ "\"\\\$TEST\"" \
+ "arg is escaped shell variable"
+ }
-run_test_same "'\\\"'" \
- "\"${bs}${bs}${bs}\"\"" \
- "'\"' is not escaped in single quote arg"
+ run_test "\$(echo foo)" \
+ "\"foo\"" \
+ "\"\\\$\\(echo\"" \
+ "arg is parameter expansion, command execution" \
+ $is_remote_with_split_args
+
+ run_test "\$((2 + 3))" \
+ "\"5\"" \
+ "\"\\\$\\(\\(2\"" \
+ "arg is parameter expansion, expression evaluation" \
+ $is_remote_with_split_args
+
+ run_test_same "\"\\a\"" \
+ "\"${bs}${bs}a\"" \
+ "retain backslash in double quote arg"
+
+ run_test_same "'\\a'" \
+ "\"${bs}${bs}a\"" \
+ "retain backslash in single quote arg"
+
+ run_test_same "\"\\\$\"" \
+ "\"\\\$\"" \
+ "'\$' can be escaped in double quote arg"
+
+ run_test_same "'\\\$'" \
+ "\"${bs}${bs}\\\$\"" \
+ "'\$' is not escaped in single quote arg"
+
+ run_test_same "\"\\`\"" \
+ "\"\\`\"" \
+ "'`' can be escaped in double quote arg"
+
+ run_test_same "'\\`'" \
+ "\"${bs}${bs}`\"" \
+ "'`' is not escaped in single quote arg"
+
+ run_test_same "\"\\\"\"" \
+ "\"${bs}\"\"" \
+ "'\"' can be escaped in double quote arg"
+
+ run_test_same "'\\\"'" \
+ "\"${bs}${bs}${bs}\"\"" \
+ "'\"' is not escaped in single quote arg"
+
+ run_test_same "\"\\\\\"" \
+ "\"${bs}${bs}\"" \
+ "'\\' can be escaped in double quote arg"
+
+ run_test_same "'\\\\'" \
+ "\"${bs}${bs}${bs}${bs}\"" \
+ "'\\' is not escaped in single quote arg"
+}
-run_test_same "\"\\\\\"" \
- "\"${bs}${bs}\"" \
- "'\\' can be escaped in double quote arg"
+run_all_tests
-run_test_same "'\\\\'" \
- "\"${bs}${bs}${bs}${bs}\"" \
- "'\\' is not escaped in single quote arg"
+# For extended-remote targets, disable the packet which passes
+# inferior arguments as a single string. This changes how the vRun
+# (extended-remote only) packet works.
+if {[target_info gdb_protocol] == "extended-remote"} {
+ with_test_prefix "single-inferior-arg disabled" {
+ save_vars { GDBFLAGS } {
+ append GDBFLAGS " -ex \"set remote single-inferior-argument-feature-packet off\""
+ run_all_tests true
+ }
+ }
+}
@@ -2725,6 +2725,8 @@ handle_query (char *own_buf, int packet_len, int *new_packet_len_p)
if (target_supports_memory_tagging ())
cs.memory_tagging_feature = true;
}
+ else if (feature == "single-inf-arg+")
+ cs.single_inferior_argument = true;
else
{
/* Move the unknown features all together. */
@@ -2854,6 +2856,9 @@ handle_query (char *own_buf, int packet_len, int *new_packet_len_p)
if (target_supports_memory_tagging ())
strcat (own_buf, ";memory-tagging+");
+ if (cs.single_inferior_argument)
+ strcat (own_buf, ";single-inf-arg+");
+
/* Reinitialize components as needed for the new connection. */
hostio_handle_new_gdb_connection ();
target_handle_new_gdb_connection ();
@@ -3438,7 +3443,21 @@ handle_v_run (char *own_buf)
else
program_path.set (new_program_name.get ());
- program_args = gdb::remote_args::join (new_argv);
+ if (cs.single_inferior_argument)
+ {
+ if (new_argv.size () > 1)
+ {
+ write_enn (own_buf);
+ return;
+ }
+ else if (new_argv.size () == 1)
+ program_args = std::string (new_argv[0]);
+ else
+ program_args.clear ();
+ }
+ else
+ program_args = gdb::remote_args::join (new_argv);
+
free_vector_argv (new_argv);
target_create_inferior (program_path.get (), program_args);
@@ -192,6 +192,11 @@ struct client_state
/* If true, memory tagging features are supported. */
bool memory_tagging_feature = false;
+ /* If true then we've agreed that the debugger will send all inferior
+ arguments as a single string. When false the debugger will attempt
+ to split the inferior arguments before sending them. */
+ bool single_inferior_argument = false;
+
};
client_state &get_client_state ();