[rfc] Fix PR 18208: update /proc/pid/coredump_filter by c code

  From: Yao Qi <yao.qi@linaro.org>

Hi,
We see some fails in gdb.base/coredump-filter.exp when we do remote
gdbserver testing, like what I did for arm/aarch64 linux testing or
run it with board file remote-gdbserver-on-localhost

 $ make check RUNTESTFLAGS='--target_board=remote-gdbserver-on-localhost coredump-filter.exp'

we find that this line in the test doesn't work as expected,

 remote_exec target "sh -c \"echo $filter_flag > /proc/$ipid/coredump_filter\""

although such pattern has been used in gdb testsuite somewhere else,
but the special thing here is that we redirect the output to
/proc/$ipid/coredump_filter on the remote target.  DejaGNU will
redirect the output from the remote target to local, and looks tcl
gets confused by these two redirection.

After trying pass different parameters to remote_exec and hacking
remote_exec/rsh_exec/local_exec, I got no success, I decide
to give up, and try to update /proc/$ipid/coredump_filter by the c
code directly.

This patch adds a c function set_coredump_filter to update
coredump_filter, and GDB calls it.

Now this test passes on boardfile unix, native-gdbserver and
remote-gdbserver-on-localhost.  It also passes my board files for arm
and aarch64 testing.

gdb/testsuite:

2015-04-24  Yao Qi  <yao.qi@linaro.org>

	PR gdb/18208
	* gdb.base/coredump-filter.c (set_coredump_filter): New function.
	* gdb.base/coredump-filter.exp (do_save_core): Call inferior
	function set_coredump_filter, and remove remote_exec call.
	Remove argument ipid.  Callers update.
	(top level): Don't get inferior's PID.
---
 gdb/testsuite/gdb.base/coredump-filter.c   | 16 ++++++++++++++++
 gdb/testsuite/gdb.base/coredump-filter.exp | 15 ++++++---------
 2 files changed, 22 insertions(+), 9 deletions(-)
  

On 04/24/2015 04:28 PM, Yao Qi wrote:

> We see some fails in gdb.base/coredump-filter.exp when we do remote
> gdbserver testing, like what I did for arm/aarch64 linux testing or
> run it with board file remote-gdbserver-on-localhost
> 
>  $ make check RUNTESTFLAGS='--target_board=remote-gdbserver-on-localhost coredump-filter.exp'
> 
> we find that this line in the test doesn't work as expected,
> 
>  remote_exec target "sh -c \"echo $filter_flag > /proc/$ipid/coredump_filter\""
> 
> although such pattern has been used in gdb testsuite somewhere else,
> but the special thing here is that we redirect the output to
> /proc/$ipid/coredump_filter on the remote target.  DejaGNU will
> redirect the output from the remote target to local, and looks tcl
> gets confused by these two redirection.

Not sure what exactly gets confused either.

> 
> After trying pass different parameters to remote_exec and hacking
> remote_exec/rsh_exec/local_exec, I got no success, I decide
> to give up, and try to update /proc/$ipid/coredump_filter by the c
> code directly.

Probably the right fix would be for dejagnu to put ''s around
the whole sh -c command in rsh_exec:

    set ret [local_exec "$RSH $rsh_useropts $hostname sh -c '$program $pargs \\; echo XYZ\\\${?}ZYX'" $inp $outp $timeout]

dunno if that would work with real rsh.  Alternatively, teach
dejagnu about a real ssh mode...

> 
> This patch adds a c function set_coredump_filter to update
> coredump_filter, and GDB calls it.
> 

This is fine with me.

> Now this test passes on boardfile unix, native-gdbserver and
> remote-gdbserver-on-localhost.  It also passes my board files for arm
> and aarch64 testing.


>      # Generate a corefile.
>      gdb_gcore_cmd "$core" "save corefile"
> @@ -146,11 +146,8 @@ if { !$core_supported } {
>      return -1
>  }
>  
> -# Get the inferior's PID.
> -set infpid ""
>  gdb_test_multiple "info inferiors" "getting inferior pid" {
> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
> -	set infpid $expect_out(1,string)
> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>      }
>      -re "Remote target.*$gdb_prompt $" {
>  	# If the target does not provide PID information (like usermode QEMU),

This "If the target does not provide PID information" check sounds
odd now.  Do we still need it?

Thanks,
Pedro Alves
  

On 05/06/2015 01:12 PM, Pedro Alves wrote:
> On 04/24/2015 04:28 PM, Yao Qi wrote:
>
>> We see some fails in gdb.base/coredump-filter.exp when we do remote
>> gdbserver testing, like what I did for arm/aarch64 linux testing or
>> run it with board file remote-gdbserver-on-localhost
>>
>>   $ make check RUNTESTFLAGS='--target_board=remote-gdbserver-on-localhost coredump-filter.exp'
>>
>> we find that this line in the test doesn't work as expected,
>>
>>   remote_exec target "sh -c \"echo $filter_flag > /proc/$ipid/coredump_filter\""
>>
>> although such pattern has been used in gdb testsuite somewhere else,
>> but the special thing here is that we redirect the output to
>> /proc/$ipid/coredump_filter on the remote target.  DejaGNU will
>> redirect the output from the remote target to local, and looks tcl
>> gets confused by these two redirection.
>
> Not sure what exactly gets confused either.
>
>>
>> After trying pass different parameters to remote_exec and hacking
>> remote_exec/rsh_exec/local_exec, I got no success, I decide
>> to give up, and try to update /proc/$ipid/coredump_filter by the c
>> code directly.
>
> Probably the right fix would be for dejagnu to put ''s around
> the whole sh -c command in rsh_exec:
>
>      set ret [local_exec "$RSH $rsh_useropts $hostname sh -c '$program $pargs \\; echo XYZ\\\${?}ZYX'" $inp $outp $timeout]
>
> dunno if that would work with real rsh.  Alternatively, teach
> dejagnu about a real ssh mode...
>
>>
>> This patch adds a c function set_coredump_filter to update
>> coredump_filter, and GDB calls it.
>>
>
> This is fine with me.
>
>> Now this test passes on boardfile unix, native-gdbserver and
>> remote-gdbserver-on-localhost.  It also passes my board files for arm
>> and aarch64 testing.
>
>
>>       # Generate a corefile.
>>       gdb_gcore_cmd "$core" "save corefile"
>> @@ -146,11 +146,8 @@ if { !$core_supported } {
>>       return -1
>>   }
>>
>> -# Get the inferior's PID.
>> -set infpid ""
>>   gdb_test_multiple "info inferiors" "getting inferior pid" {
>> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
>> -	set infpid $expect_out(1,string)
>> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>>       }
>>       -re "Remote target.*$gdb_prompt $" {
>>   	# If the target does not provide PID information (like usermode QEMU),
>
> This "If the target does not provide PID information" check sounds
> odd now.  Do we still need it?

If we're not dealing with PID's, i don't think so.
  

Luis Machado <lgustavo@codesourcery.com> writes:

>>> -# Get the inferior's PID.
>>> -set infpid ""
>>>   gdb_test_multiple "info inferiors" "getting inferior pid" {
>>> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
>>> -	set infpid $expect_out(1,string)
>>> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>>>       }
>>>       -re "Remote target.*$gdb_prompt $" {
>>>   	# If the target does not provide PID information (like usermode QEMU),
>>
>> This "If the target does not provide PID information" check sounds
>> odd now.  Do we still need it?
>
> If we're not dealing with PID's, i don't think so.

At the very start, I removed this block, but I recall that this block is
used as a guard for usermode QEMU which doesn't provide PID
information.  With this patch applied, we'll access
/proc/self/coredump_filter, but I am afraid it doesn't work as expected
on usermode QEMU, because usermode QEMU just intercepts few /proc
accesses and pass most of them through the host linux.  Accessing
/proc/QEMU_PID/coredump_filter isn't what we want in this test, so I
think it's better to skip the test for usermode QEMU.

Of course, I don't mind removing this block.  Luis, could you try this
patch and remove this block, see whether it causes fails on usermode QEMU?
  

On 05/07/2015 06:05 AM, Yao Qi wrote:
> Luis Machado <lgustavo@codesourcery.com> writes:
>
>>>> -# Get the inferior's PID.
>>>> -set infpid ""
>>>>    gdb_test_multiple "info inferiors" "getting inferior pid" {
>>>> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
>>>> -	set infpid $expect_out(1,string)
>>>> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>>>>        }
>>>>        -re "Remote target.*$gdb_prompt $" {
>>>>    	# If the target does not provide PID information (like usermode QEMU),
>>>
>>> This "If the target does not provide PID information" check sounds
>>> odd now.  Do we still need it?
>>
>> If we're not dealing with PID's, i don't think so.
>
> At the very start, I removed this block, but I recall that this block is
> used as a guard for usermode QEMU which doesn't provide PID
> information.  With this patch applied, we'll access
> /proc/self/coredump_filter, but I am afraid it doesn't work as expected
> on usermode QEMU, because usermode QEMU just intercepts few /proc
> accesses and pass most of them through the host linux.  Accessing
> /proc/QEMU_PID/coredump_filter isn't what we want in this test, so I
> think it's better to skip the test for usermode QEMU.
>
> Of course, I don't mind removing this block.  Luis, could you try this
> patch and remove this block, see whether it causes fails on usermode QEMU?
>

Yeah, that sounds problematic. I'll give it a try and will let you know.
  

On 05/07/2015 07:44 AM, Luis Machado wrote:
> On 05/07/2015 06:05 AM, Yao Qi wrote:
>> Luis Machado <lgustavo@codesourcery.com> writes:
>>
>>>>> -# Get the inferior's PID.
>>>>> -set infpid ""
>>>>>    gdb_test_multiple "info inferiors" "getting inferior pid" {
>>>>> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
>>>>> -    set infpid $expect_out(1,string)
>>>>> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>>>>>        }
>>>>>        -re "Remote target.*$gdb_prompt $" {
>>>>>        # If the target does not provide PID information (like
>>>>> usermode QEMU),
>>>>
>>>> This "If the target does not provide PID information" check sounds
>>>> odd now.  Do we still need it?
>>>
>>> If we're not dealing with PID's, i don't think so.
>>
>> At the very start, I removed this block, but I recall that this block is
>> used as a guard for usermode QEMU which doesn't provide PID
>> information.  With this patch applied, we'll access
>> /proc/self/coredump_filter, but I am afraid it doesn't work as expected
>> on usermode QEMU, because usermode QEMU just intercepts few /proc
>> accesses and pass most of them through the host linux.  Accessing
>> /proc/QEMU_PID/coredump_filter isn't what we want in this test, so I
>> think it's better to skip the test for usermode QEMU.
>>
>> Of course, I don't mind removing this block.  Luis, could you try this
>> patch and remove this block, see whether it causes fails on usermode
>> QEMU?
>>
>
> Yeah, that sounds problematic. I'll give it a try and will let you know.

Removing that conditional block i get 14 FAIL's, so it doesn't look like 
this test is suited for usermode QEMU.

Besides that, without the conditional block i also see a problem with 
the existing testcase using coredump_var_addr as an array but declaring 
it as a regular variable. It causes a number of errors. Did you notice 
that? Maybe it is specific to a certain version of expect.

# Obtain the address of each variable that will be checked on each
# case.
set coredump_var_addr ""
foreach item $all_anon_corefiles {
     foreach name [list [lindex $item 3] [lindex $item 4]] {
         set test "print/x $name"
         gdb_test_multiple $test $test {
             -re " = \($hex\)\r\n$gdb_prompt $" {
                 set coredump_var_addr($name) $expect_out(1,string)
             }
         }
     }
}
  

On Wednesday, May 06 2015, Pedro Alves wrote:

>> After trying pass different parameters to remote_exec and hacking
>> remote_exec/rsh_exec/local_exec, I got no success, I decide
>> to give up, and try to update /proc/$ipid/coredump_filter by the c
>> code directly.
>
> Probably the right fix would be for dejagnu to put ''s around
> the whole sh -c command in rsh_exec:
>
>     set ret [local_exec "$RSH $rsh_useropts $hostname sh -c '$program $pargs \\; echo XYZ\\\${?}ZYX'" $inp $outp $timeout]
>
> dunno if that would work with real rsh.  Alternatively, teach
> dejagnu about a real ssh mode...

FWIW, I managed to solve the problem by hacking
/usr/share/dejagnu/rsh.exp:rsh_exec and doing:

  set ret [local_exec "$RSH $rsh_useropts $hostname \"sh -c '$program $pargs \\; echo XYZ\\\${?}ZYX'\"" $inp $outp $timeout]

Then I found this message by Pedro.

But only doing this is not enough: because of the quote-hell, the
coredump-filter.exp needs to be adjusted to:

  remote_exec target "echo $filter_flag > /proc/$ipid/coredump_filter"

I.e., remove the "sh -c" and the extra quotes.  After that, everything
passes.

Since the full solution involves hacking dejagnu, then I agree that a
most desirable solution is to hack the C source file as Yao did.

Thanks,
  

Luis Machado <lgustavo@codesourcery.com> writes:

> Removing that conditional block i get 14 FAIL's, so it doesn't look
> like this test is suited for usermode QEMU.

I'll leave that block (checking PID) there then...
Patch is pushed in.
  

On 05/07/2015 03:01 PM, Luis Machado wrote:
> On 05/07/2015 07:44 AM, Luis Machado wrote:
>> On 05/07/2015 06:05 AM, Yao Qi wrote:
>>> Luis Machado <lgustavo@codesourcery.com> writes:
>>>
>>>>>> -# Get the inferior's PID.
>>>>>> -set infpid ""
>>>>>>    gdb_test_multiple "info inferiors" "getting inferior pid" {
>>>>>> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
>>>>>> -    set infpid $expect_out(1,string)
>>>>>> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>>>>>>        }
>>>>>>        -re "Remote target.*$gdb_prompt $" {
>>>>>>        # If the target does not provide PID information (like
>>>>>> usermode QEMU),
>>>>>
>>>>> This "If the target does not provide PID information" check sounds
>>>>> odd now.  Do we still need it?
>>>>
>>>> If we're not dealing with PID's, i don't think so.
>>>
>>> At the very start, I removed this block, but I recall that this block is
>>> used as a guard for usermode QEMU which doesn't provide PID
>>> information.  With this patch applied, we'll access
>>> /proc/self/coredump_filter, but I am afraid it doesn't work as expected
>>> on usermode QEMU, because usermode QEMU just intercepts few /proc
>>> accesses and pass most of them through the host linux.  Accessing
>>> /proc/QEMU_PID/coredump_filter isn't what we want in this test, so I
>>> think it's better to skip the test for usermode QEMU.
>>>
>>> Of course, I don't mind removing this block.  Luis, could you try this
>>> patch and remove this block, see whether it causes fails on usermode
>>> QEMU?
>>>
>>
>> Yeah, that sounds problematic. I'll give it a try and will let you know.
> 
> Removing that conditional block i get 14 FAIL's, so it doesn't look like 
> this test is suited for usermode QEMU.

But what does gdb.log show?

With usermode QEMU, the program and qemu are the same process, thus
have the same PID.  I just tried loading up the test's probably (manually
compiled) under F20's qemu-arm, generating a core with gcore, and then
loading the core back into gdb, which worked.

I didn't test beyond that as I don't have a usermode qemu board
file handy (it'd be nice to have one in testsuite/boards/).

I'm not immediately seeing the fundamental reason this shouldn't
have worked, and we may be hiding a bug instead.

Thanks,
Pedro Alves
  

On 05/08/2015 11:40 AM, Pedro Alves wrote:
> On 05/07/2015 03:01 PM, Luis Machado wrote:
>> On 05/07/2015 07:44 AM, Luis Machado wrote:
>>> On 05/07/2015 06:05 AM, Yao Qi wrote:
>>>> Luis Machado <lgustavo@codesourcery.com> writes:
>>>>
>>>>>>> -# Get the inferior's PID.
>>>>>>> -set infpid ""
>>>>>>>     gdb_test_multiple "info inferiors" "getting inferior pid" {
>>>>>>> -    -re "process \($decimal\).*\r\n$gdb_prompt $" {
>>>>>>> -    set infpid $expect_out(1,string)
>>>>>>> +    -re "process $decimal.*\r\n$gdb_prompt $" {
>>>>>>>         }
>>>>>>>         -re "Remote target.*$gdb_prompt $" {
>>>>>>>         # If the target does not provide PID information (like
>>>>>>> usermode QEMU),
>>>>>>
>>>>>> This "If the target does not provide PID information" check sounds
>>>>>> odd now.  Do we still need it?
>>>>>
>>>>> If we're not dealing with PID's, i don't think so.
>>>>
>>>> At the very start, I removed this block, but I recall that this block is
>>>> used as a guard for usermode QEMU which doesn't provide PID
>>>> information.  With this patch applied, we'll access
>>>> /proc/self/coredump_filter, but I am afraid it doesn't work as expected
>>>> on usermode QEMU, because usermode QEMU just intercepts few /proc
>>>> accesses and pass most of them through the host linux.  Accessing
>>>> /proc/QEMU_PID/coredump_filter isn't what we want in this test, so I
>>>> think it's better to skip the test for usermode QEMU.
>>>>
>>>> Of course, I don't mind removing this block.  Luis, could you try this
>>>> patch and remove this block, see whether it causes fails on usermode
>>>> QEMU?
>>>>
>>>
>>> Yeah, that sounds problematic. I'll give it a try and will let you know.
>>
>> Removing that conditional block i get 14 FAIL's, so it doesn't look like
>> this test is suited for usermode QEMU.
>
> But what does gdb.log show?
>
> With usermode QEMU, the program and qemu are the same process, thus
> have the same PID.  I just tried loading up the test's probably (manually
> compiled) under F20's qemu-arm, generating a core with gcore, and then
> loading the core back into gdb, which worked.
>
> I didn't test beyond that as I don't have a usermode qemu board
> file handy (it'd be nice to have one in testsuite/boards/).
>
> I'm not immediately seeing the fundamental reason this shouldn't
> have worked, and we may be hiding a bug instead.

I'll have it reproduced again and will inspect the log. I recall seeing 
nonsense and random PC addresses that didn't point to proper symbols.