[PATCH] Harden gdb.arch/aarch64-pauth.exp and fix a failure

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

[PATCH] Harden gdb.arch/aarch64-pauth.exp and fix a failure

Luis Machado-2
When running this testcase against a QEMU with PAC support, i noticed we
were failing to recognize the additional [PAC] that is emitted in the
backtrace, resulting in this failure:

FAIL: gdb.arch/aarch64-pauth.exp: backtrace

I've made the test use multi_line to make the pattern more clear.

Tested against aarch64-linux-gnu with and without PAC support.

gdb/testsuite/ChangeLog:

2020-02-21  Luis Machado  <[hidden email]>

        * gdb.arch/aarch64-pauth.exp: Recognize optional PAC output.

Signed-off-by: Luis Machado <[hidden email]>
---
 gdb/testsuite/gdb.arch/aarch64-pauth.exp | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/gdb/testsuite/gdb.arch/aarch64-pauth.exp b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
index 816e58be44..e6d25c5d97 100644
--- a/gdb/testsuite/gdb.arch/aarch64-pauth.exp
+++ b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
@@ -40,4 +40,9 @@ gdb_breakpoint [ gdb_get_line_number "break here" ]
 gdb_continue_to_breakpoint "break here" ".*break here.*"
 
 # Ensure we can get a full backtrace, despite the address signing.
-gdb_test "bt" "^bt\r\n#0 +bar *\\(b=9\\) +at.*\r\n#1 +0x\[0-9a-f\]* +in +foo \\(a=5\\).*\r\n#2 +0x\[0-9a-f\]* +in +main \\(\\).*" "backtrace"
+gdb_test "bt" \
+    [multi_line \
+ "#0\[ \t\]*bar \\(b=9\\) at \[^\r\n\]+" \
+ "#1\[ \t\]*$hex (\\\[PAC\\\] )?in foo \\(a=5\\) at \[^\r\n\]+" \
+ "#2\[ \t\]*$hex (\\\[PAC\\\] )?in main \\(\\) at \[^\r\n\]+" ] \
+    "backtrace"
--
2.17.1

Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] Harden gdb.arch/aarch64-pauth.exp and fix a failure

Kevin Buettner
On Fri, 21 Feb 2020 18:06:28 -0300
Luis Machado <[hidden email]> wrote:

> When running this testcase against a QEMU with PAC support, i noticed we
> were failing to recognize the additional [PAC] that is emitted in the
> backtrace, resulting in this failure:
>
> FAIL: gdb.arch/aarch64-pauth.exp: backtrace
>
> I've made the test use multi_line to make the pattern more clear.
>
> Tested against aarch64-linux-gnu with and without PAC support.
>
> gdb/testsuite/ChangeLog:
>
> 2020-02-21  Luis Machado  <[hidden email]>
>
> * gdb.arch/aarch64-pauth.exp: Recognize optional PAC output.
>
> Signed-off-by: Luis Machado <[hidden email]>
> ---
>  gdb/testsuite/gdb.arch/aarch64-pauth.exp | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/gdb/testsuite/gdb.arch/aarch64-pauth.exp b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
> index 816e58be44..e6d25c5d97 100644
> --- a/gdb/testsuite/gdb.arch/aarch64-pauth.exp
> +++ b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
> @@ -40,4 +40,9 @@ gdb_breakpoint [ gdb_get_line_number "break here" ]
>  gdb_continue_to_breakpoint "break here" ".*break here.*"
>  
>  # Ensure we can get a full backtrace, despite the address signing.
> -gdb_test "bt" "^bt\r\n#0 +bar *\\(b=9\\) +at.*\r\n#1 +0x\[0-9a-f\]* +in +foo \\(a=5\\).*\r\n#2 +0x\[0-9a-f\]* +in +main \\(\\).*" "backtrace"
> +gdb_test "bt" \
> +    [multi_line \
> + "#0\[ \t\]*bar \\(b=9\\) at \[^\r\n\]+" \
> + "#1\[ \t\]*$hex (\\\[PAC\\\] )?in foo \\(a=5\\) at \[^\r\n\]+" \
> + "#2\[ \t\]*$hex (\\\[PAC\\\] )?in main \\(\\) at \[^\r\n\]+" ] \
> +    "backtrace"

LGTM except for the \[ \t\]* parts of it.  I think we actually want +
instead of * to ensure that there's whitespace separating the frame
number from the hex value.

Kevin

Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] Harden gdb.arch/aarch64-pauth.exp and fix a failure

Luis Machado-2
On 2/22/20 1:20 AM, Kevin Buettner wrote:

> On Fri, 21 Feb 2020 18:06:28 -0300
> Luis Machado <[hidden email]> wrote:
>
>> When running this testcase against a QEMU with PAC support, i noticed we
>> were failing to recognize the additional [PAC] that is emitted in the
>> backtrace, resulting in this failure:
>>
>> FAIL: gdb.arch/aarch64-pauth.exp: backtrace
>>
>> I've made the test use multi_line to make the pattern more clear.
>>
>> Tested against aarch64-linux-gnu with and without PAC support.
>>
>> gdb/testsuite/ChangeLog:
>>
>> 2020-02-21  Luis Machado  <[hidden email]>
>>
>> * gdb.arch/aarch64-pauth.exp: Recognize optional PAC output.
>>
>> Signed-off-by: Luis Machado <[hidden email]>
>> ---
>>   gdb/testsuite/gdb.arch/aarch64-pauth.exp | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/gdb/testsuite/gdb.arch/aarch64-pauth.exp b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
>> index 816e58be44..e6d25c5d97 100644
>> --- a/gdb/testsuite/gdb.arch/aarch64-pauth.exp
>> +++ b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
>> @@ -40,4 +40,9 @@ gdb_breakpoint [ gdb_get_line_number "break here" ]
>>   gdb_continue_to_breakpoint "break here" ".*break here.*"
>>  
>>   # Ensure we can get a full backtrace, despite the address signing.
>> -gdb_test "bt" "^bt\r\n#0 +bar *\\(b=9\\) +at.*\r\n#1 +0x\[0-9a-f\]* +in +foo \\(a=5\\).*\r\n#2 +0x\[0-9a-f\]* +in +main \\(\\).*" "backtrace"
>> +gdb_test "bt" \
>> +    [multi_line \
>> + "#0\[ \t\]*bar \\(b=9\\) at \[^\r\n\]+" \
>> + "#1\[ \t\]*$hex (\\\[PAC\\\] )?in foo \\(a=5\\) at \[^\r\n\]+" \
>> + "#2\[ \t\]*$hex (\\\[PAC\\\] )?in main \\(\\) at \[^\r\n\]+" ] \
>> +    "backtrace"
>
> LGTM except for the \[ \t\]* parts of it.  I think we actually want +
> instead of * to ensure that there's whitespace separating the frame
> number from the hex value.
>
> Kevin
>

Indeed. Fixed now.

Thanks for the review.
Reply | Threaded
Open this post in threaded view
|

Re: [PATCH] Harden gdb.arch/aarch64-pauth.exp and fix a failure

Luis Machado-2
In reply to this post by Kevin Buettner
On 2/22/20 1:20 AM, Kevin Buettner wrote:

> On Fri, 21 Feb 2020 18:06:28 -0300
> Luis Machado <[hidden email]> wrote:
>
>> When running this testcase against a QEMU with PAC support, i noticed we
>> were failing to recognize the additional [PAC] that is emitted in the
>> backtrace, resulting in this failure:
>>
>> FAIL: gdb.arch/aarch64-pauth.exp: backtrace
>>
>> I've made the test use multi_line to make the pattern more clear.
>>
>> Tested against aarch64-linux-gnu with and without PAC support.
>>
>> gdb/testsuite/ChangeLog:
>>
>> 2020-02-21  Luis Machado  <[hidden email]>
>>
>> * gdb.arch/aarch64-pauth.exp: Recognize optional PAC output.
>>
>> Signed-off-by: Luis Machado <[hidden email]>
>> ---
>>   gdb/testsuite/gdb.arch/aarch64-pauth.exp | 7 ++++++-
>>   1 file changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/gdb/testsuite/gdb.arch/aarch64-pauth.exp b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
>> index 816e58be44..e6d25c5d97 100644
>> --- a/gdb/testsuite/gdb.arch/aarch64-pauth.exp
>> +++ b/gdb/testsuite/gdb.arch/aarch64-pauth.exp
>> @@ -40,4 +40,9 @@ gdb_breakpoint [ gdb_get_line_number "break here" ]
>>   gdb_continue_to_breakpoint "break here" ".*break here.*"
>>  
>>   # Ensure we can get a full backtrace, despite the address signing.
>> -gdb_test "bt" "^bt\r\n#0 +bar *\\(b=9\\) +at.*\r\n#1 +0x\[0-9a-f\]* +in +foo \\(a=5\\).*\r\n#2 +0x\[0-9a-f\]* +in +main \\(\\).*" "backtrace"
>> +gdb_test "bt" \
>> +    [multi_line \
>> + "#0\[ \t\]*bar \\(b=9\\) at \[^\r\n\]+" \
>> + "#1\[ \t\]*$hex (\\\[PAC\\\] )?in foo \\(a=5\\) at \[^\r\n\]+" \
>> + "#2\[ \t\]*$hex (\\\[PAC\\\] )?in main \\(\\) at \[^\r\n\]+" ] \
>> +    "backtrace"
>
> LGTM except for the \[ \t\]* parts of it.  I think we actually want +
> instead of * to ensure that there's whitespace separating the frame
> number from the hex value.
>
> Kevin
>

Fixed and pushed. Thanks!

I'll also push a fix to gdb.base/backtrace.exp, since it also uses *
instead of + for the following tests:

# Backtrace with the default options.
gdb_test "bt" \
     [multi_line \
          "#0\[ \t\]*baz \\(\\) at \[^\r\n\]+" \
          "#1\[ \t\]*$hex in bar \\(\\) at \[^\r\n\]+" \
          "#2\[ \t\]*$hex in foo \\(\\) at \[^\r\n\]+" \
          "#3\[ \t\]*$hex in main \\(\\) at \[^\r\n\]+" ]

# Backtrace with 'set disassemble-next-line on'.  This shouldn't make
# any difference to the backtrace.
gdb_test "with disassemble-next-line on -- bt" \
     [multi_line \
          "#0\[ \t\]*baz \\(\\) at \[^\r\n\]+" \
          "#1\[ \t\]*$hex in bar \\(\\) at \[^\r\n\]+" \
          "#2\[ \t\]*$hex in foo \\(\\) at \[^\r\n\]+" \
          "#3\[ \t\]*$hex in main \\(\\) at \[^\r\n\]+" ]