[RFC] Have an option to tell GDB to detect and possibly handle mismatched exec-files.

Message ID 20190318225822.4820-1-philippe.waroquiers@skynet.be
State New
Headers show
Series
  • [RFC] Have an option to tell GDB to detect and possibly handle mismatched exec-files.
Related show

Commit Message

Philippe Waroquiers March 18, 2019, 10:58 p.m.
A recurrent problem with GDB is that GDB uses the wrong exec-file
when using the attach/detach commands sucessfully.
Also, in case the user specifies a file on the command line but attaches
to the wrong PID, this error is not made visible and gives a not user
understandable behaviour.

For example:
  $ gdb
  ...
  (gdb) atta 2682
  Attaching to process 2682
  [New LWP 2683]
  [New LWP 2684]
  [New LWP 2685]
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  0x00007f5ff829f603 in select () at ../sysdeps/unix/syscall-template.S:84
  84	../sysdeps/unix/syscall-template.S: No such file or directory.
  (gdb) det
  Detaching from program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 2682
  [Inferior 1 (process 2682) detached]
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  Reading symbols from /lib64/ld-linux-x86-64.so.2...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in ?? ()
  (gdb) bt
  #0  0x00007f43c23478a0 in ?? ()
  #1  0x0000558909e3ad91 in ?? ()
  #2  0x0000202962646700 in ?? ()
  #3  0x00007ffc69c74e70 in ?? ()
  #4  0x000055890c1d2350 in ?? ()
  #5  0x0000000000000000 in ?? ()
  (gdb)

The second attach has kept the executable of the first attach.
(in this case, 31069 is the PID of a GDB, that has nothing to do
with the first determined 'sleepers' executable).

Similarly, if specifying an executable, but attaching to a wrong pid,
we get:

  gdb /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  ...
  Reading symbols from /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers...
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  Reading symbols from /lib64/ld-linux-x86-64.so.2...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in ?? ()
  (gdb) bt
  #0  0x00007f43c23478a0 in ?? ()
  #1  0x0000558909e3ad91 in ?? ()
  #2  0x0000202962646700 in ?? ()
  #3  0x00007ffc69c74e70 in ?? ()
  #4  0x000055890c1d2350 in ?? ()
  #5  0x0000000000000000 in ?? ()
  (gdb)

And it is unclear to the user what has happened/what is going wrong.

This patch series implements a new option:
  (gdb) apropos exec-file-mismatch
  set exec-file-mismatch -- Set exec-file-mismatch handling
  show exec-file-mismatch -- Show exec-file-mismatch handling
  (gdb) help set exec-file-mismatch
  Set exec-file-mismatch handling.   (reload/warn/off)
  (gdb)

"reload" means: in case of mismatch between the current exec-file
and the automatically determined exec-file of the PID we are attaching to,
give a warning to the user and reload the automatically determined exec-file.
This reload might be subject to a question to the user.

"warn" means: in case of mismatch, just give a warning to the user.

"off" means: do not check for mismatch.

This fixes PR gdb/17626.
There was a previous trial to fix this PR.
See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html
This trial was however only fixing the problem for the automatically
determined executable files when doing attach.
It was differentiating the 'user specified executable files' ("sticky")
from the executable files automatically found by GDB.
But such user specified sticky executables are in most cases due
to a wrong manipulation by the user, giving unexpected results
such as backtrace showing no function like in the above example.

This patch ensures that whenever a process executable can be
determined, that the user is warned if there is a mismatch.

The same tests as above then give:

  (gdb) atta 2682
  Attaching to process 2682
  [New LWP 2683]
  [New LWP 2684]
  [New LWP 2685]
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  0x00007f5ff829f603 in select () at ../sysdeps/unix/syscall-template.S:84
  84	../sysdeps/unix/syscall-template.S: No such file or directory.
  (gdb) det
  Detaching from program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 2682
  [Inferior 1 (process 2682) detached]
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
  exec-file-mismatch handling is currently "reload"
  Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
  Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...
  Setting up the environment for debugging gdb.
  ...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in __poll_nocancel () at ../sysdeps/unix/syscall-template.S:84
  84	../sysdeps/unix/syscall-template.S: No such file or directory.
  (top-gdb) bt
  During symbol reading: incomplete CFI data; unspecified registers (e.g., rax) at 0x7f43c23478ad
  During symbol reading: unsupported tag: 'DW_TAG_unspecified_type'
  During symbol reading: cannot get low and high bounds for subprogram DIE at 0x12282a7
  During symbol reading: Child DIE 0x12288ba and its abstract origin 0x1228b26 have different parents
  During symbol reading: DW_AT_call_target target DIE has invalid low pc, for referencing DIE 0x1229540 [in module /bd/home/philippe/gdb/git/build_fixes/gdb/gdb]
  #0  0x00007f43c23478a0 in __poll_nocancel () at ../sysdeps/unix/syscall-template.S:84
  #1  0x0000558909e3ad91 in poll (__timeout=-1, __nfds=<optimized out>, __fds=<optimized out>) at /usr/include/x86_64-linux-gnu/bits/poll2.h:46
  #2  gdb_wait_for_event (block=block@entry=1) at ../../fixes/gdb/event-loop.c:772
  #3  0x0000558909e3aef4 in gdb_do_one_event () at ../../fixes/gdb/event-loop.c:347
  #4  0x0000558909e3b085 in gdb_do_one_event () at ../../fixes/gdb/common/common-exceptions.h:219
  #5  start_event_loop () at ../../fixes/gdb/event-loop.c:371
  During symbol reading: Member function "~_Sp_counted_base" (offset 0x1c69bf7) is virtual but the vtable offset is not specified
  During symbol reading: Multiple children of DIE 0x1c8f5a0 refer to DIE 0x1c8f0ee as their abstract origin
  #6  0x0000558909ed3b78 in captured_command_loop () at ../../fixes/gdb/main.c:331
  #7  0x0000558909ed4b6d in captured_main (data=<optimized out>) at ../../fixes/gdb/main.c:1174
  #8  gdb_main (args=<optimized out>) at ../../fixes/gdb/main.c:1190
  #9  0x0000558909c1e9a8 in main (argc=<optimized out>, argv=<optimized out>) at ../../fixes/gdb/gdb.c:32
  (top-gdb)

  gdb /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepersReading symbols from /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers...
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
  exec-file-mismatch handling is currently "reload"
  Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
  Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...
  Setting up the environment for debugging gdb.
  ....

In other words, it now works as intuitively expected by the user.
If ever the user gave the correct executable on the command line,
then attached to the wrong pid, then confirmed loading the wrong executable,
the user can simply fix this by detaching, and attaching to the correct pid,
GDB will then tell again to the user that the exec-file might better
be reloaded.

The default value of "reload" is chosen instead of e.g. "warn" as in most
cases, switching of executable will be the correct action,
and in any case, the user can decide to not load the executable.

Tested on debian/amd64, no regression.

Still missing:
  ChangeLog
  NEWS
  user manual
---
 gdb/exec.c    | 104 ++++++++++++++++++++++++++++++++++++++++++++++++++
 gdb/gdbcore.h |   5 +++
 gdb/infcmd.c  |   3 ++
 gdb/remote.c  |   3 ++
 4 files changed, 115 insertions(+)

-- 
2.20.1

Comments

Tom Tromey March 19, 2019, 4:43 p.m. | #1
>>>>> "Philippe" == Philippe Waroquiers <philippe.waroquiers@skynet.be> writes:


Philippe> A recurrent problem with GDB is that GDB uses the wrong exec-file
Philippe> when using the attach/detach commands sucessfully.
Philippe> Also, in case the user specifies a file on the command line but attaches
Philippe> to the wrong PID, this error is not made visible and gives a not user
Philippe> understandable behaviour.

I've been bitten any number of times by this.  So, it would be great to
have this fixed.

Philippe> There was a previous trial to fix this PR.
Philippe> See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html
Philippe> This trial was however only fixing the problem for the automatically
Philippe> determined executable files when doing attach.

There was also this series:

https://sourceware.org/ml/gdb-patches/2014-03/msg00476.html

... which seemingly went through at least 4 rounds of review and then
did not land for some reason.

Before approving anything I would like to understand why neither of
these earlier attempts went in.

The build-id approach is appealing because (especially since the demise
of prelink) it seems that gdb could simply trust it and abandon
non-matching symbol files, without needing a "user supplied" bit.

Tom
Philippe Waroquiers March 20, 2019, 6:34 p.m. | #2
On Tue, 2019-03-19 at 10:43 -0600, Tom Tromey wrote:
> > > > > > "Philippe" == Philippe Waroquiers <philippe.waroquiers@skynet.be> writes:

> 

> Philippe> A recurrent problem with GDB is that GDB uses the wrong exec-file

> Philippe> when using the attach/detach commands sucessfully.

> Philippe> Also, in case the user specifies a file on the command line but attaches

> Philippe> to the wrong PID, this error is not made visible and gives a not user

> Philippe> understandable behaviour.

> 

> I've been bitten any number of times by this.  So, it would be great to

> have this fixed.

> 

> Philippe> There was a previous trial to fix this PR.

> Philippe> See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html

> Philippe> This trial was however only fixing the problem for the automatically

> Philippe> determined executable files when doing attach.

I have re-read all the exchanges for the above fix trial.
It ended up with the above msg00118 with Pedro asking 'WDYT?',
and it looks like everybody was thinking nothing.
So, after more than 4 years, here is some feedback :).

IMO, the distinction between auto-discovered and user-specified
is artificial: if the user will debug a PID with what looks like a
wrong executable specified by the user, GDB should not silently
use the user provided file, as keeping it is very probably
wrong (at least in the cases I have seen, it was always me
doing an error, and the patch below will make the error visible
and easy to fix) :
The patch proposed below will indicate that there is very probably
something weird, and then the user can change the setting and/or
decide to not load the exec file discovered from the PID.

Also, the patch above makes GDB unconditionally switch of exec file
for auto-discovered
files.
If ever the user really want to do something 'unusual', no way.
Again, the patch
below allows the user to control the behaviour
if really something unusual is
desired/needed.

> 

> There was also this series:

> 

> https://sourceware.org/ml/gdb-patches/2014-03/msg00476.html

> 

> ... which seemingly went through at least 4 rounds of review and then

> did not land for some reason.

After digging in the mail archive, it looks like the last state
of this mail thread was the v5 done by Jan Kratochvil.
I found no reaction to this v5.  No idea why it stalled.

That being said, I think that this last series does something different.
IIUC, this patch series ensures that when GDB receives the pathname
of a file to use (typically from a gdbserver), that GDB will use the
build-id to decide to *not* load the debug info, but this does not
lead to 'switch' to another correct exec-file.

In other words, this v5 ensures that GDB does not load wrong
debug info for a file GDB has decided to use as exec-file (or library).
It does not detect or report that a wrong file is selected for a PID.

In that sense, the patch proposed below is not contradictory with this
last patch v5: once GDB has decided to use a certain exec-file,
GDB today checks e.g. that the file has not changed of time stamp
(otherwise it re-reads the file).  As I understand, the build-id is just
a more sophisticated way to detect if the file loaded by the PID
is consistent with either the reported file to use by GDB and/or with
the separate debug info GDB finds on the host side.

> 

> Before approving anything I would like to understand why neither of

> these earlier attempts went in.

> 

> The build-id approach is appealing because (especially since the demise

> of prelink) it seems that gdb could simply trust it and abandon

> non-matching symbol files, without needing a "user supplied" bit.


In summary, I think that the patch below is another approach
than the 'sticky user executable' concept, but does not aim at
solving the problem of an exec-file or a remote exec-file
inconsistent with some GDB host debug info.


Philippe
Philippe Waroquiers April 3, 2019, 8:24 p.m. | #3
Pedro/Jan,
Some feedback ?

Thanks

Philippe


On Wed, 2019-03-20 at 19:34 +0100, Philippe Waroquiers wrote:
> On Tue, 2019-03-19 at 10:43 -0600, Tom Tromey wrote:

> > > > > > > "Philippe" == Philippe Waroquiers <philippe.waroquiers@skynet.be> writes:

> > 

> > Philippe> A recurrent problem with GDB is that GDB uses the wrong exec-file

> > Philippe> when using the attach/detach commands sucessfully.

> > Philippe> Also, in case the user specifies a file on the command line but attaches

> > Philippe> to the wrong PID, this error is not made visible and gives a not user

> > Philippe> understandable behaviour.

> > 

> > I've been bitten any number of times by this.  So, it would be great to

> > have this fixed.

> > 

> > Philippe> There was a previous trial to fix this PR.

> > Philippe> See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html

> > Philippe> This trial was however only fixing the problem for the automatically

> > Philippe> determined executable files when doing attach.

> 

> I have re-read all the exchanges for the above fix trial.

> It ended up with the above msg00118 with Pedro asking 'WDYT?',

> and it looks like everybody was thinking nothing.

> So, after more than 4 years, here is some feedback :).

> 

> IMO, the distinction between auto-discovered and user-specified

> is artificial: if the user will debug a PID with what looks like a

> wrong executable specified by the user, GDB should not silently

> use the user provided file, as keeping it is very probably

> wrong (at least in the cases I have seen, it was always me

> doing an error, and the patch below will make the error visible

> and easy to fix) :

> The patch proposed below will indicate that there is very probably

> something weird, and then the user can change the setting and/or

> decide to not load the exec file discovered from the PID.

> 

> Also, the patch above makes GDB unconditionally switch of exec file

> for auto-discovered

> files.

> If ever the user really want to do something 'unusual', no way.

> Again, the patch

> below allows the user to control the behaviour

> if really something unusual is

> desired/needed.

> 

> > 

> > There was also this series:

> > 

> > https://sourceware.org/ml/gdb-patches/2014-03/msg00476.html

> > 

> > ... which seemingly went through at least 4 rounds of review and then

> > did not land for some reason.

> 

> After digging in the mail archive, it looks like the last state

> of this mail thread was the v5 done by Jan Kratochvil.

> I found no reaction to this v5.  No idea why it stalled.

> 

> That being said, I think that this last series does something different.

> IIUC, this patch series ensures that when GDB receives the pathname

> of a file to use (typically from a gdbserver), that GDB will use the

> build-id to decide to *not* load the debug info, but this does not

> lead to 'switch' to another correct exec-file.

> 

> In other words, this v5 ensures that GDB does not load wrong

> debug info for a file GDB has decided to use as exec-file (or library).

> It does not detect or report that a wrong file is selected for a PID.

> 

> In that sense, the patch proposed below is not contradictory with this

> last patch v5: once GDB has decided to use a certain exec-file,

> GDB today checks e.g. that the file has not changed of time stamp

> (otherwise it re-reads the file).  As I understand, the build-id is just

> a more sophisticated way to detect if the file loaded by the PID

> is consistent with either the reported file to use by GDB and/or with

> the separate debug info GDB finds on the host side.

> 

> > 

> > Before approving anything I would like to understand why neither of

> > these earlier attempts went in.

> > 

> > The build-id approach is appealing because (especially since the demise

> > of prelink) it seems that gdb could simply trust it and abandon

> > non-matching symbol files, without needing a "user supplied" bit.

> 

> In summary, I think that the patch below is another approach

> than the 'sticky user executable' concept, but does not aim at

> solving the problem of an exec-file or a remote exec-file

> inconsistent with some GDB host debug info.

> 

> 

> Philippe

> 

> 

>
Philippe Waroquiers April 10, 2019, 12:09 p.m. | #4
Some feedback ?
(or should I just go on, finish the work to submit an RFA?)
Thanks
Philippe

NB: Jan told me this was a long time ago, and could not give
much comments, as he is not working on GDB anymore.

On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:
> Pedro/Jan,

> Some feedback ?

> 

> Thanks

> 

> Philippe

> 

> 

> On Wed, 2019-03-20 at 19:34 +0100, Philippe Waroquiers wrote:

> > On Tue, 2019-03-19 at 10:43 -0600, Tom Tromey wrote:

> > > > > > > > "Philippe" == Philippe Waroquiers <philippe.waroquiers@skynet.be> writes:

> > > 

> > > Philippe> A recurrent problem with GDB is that GDB uses the wrong exec-file

> > > Philippe> when using the attach/detach commands sucessfully.

> > > Philippe> Also, in case the user specifies a file on the command line but attaches

> > > Philippe> to the wrong PID, this error is not made visible and gives a not user

> > > Philippe> understandable behaviour.

> > > 

> > > I've been bitten any number of times by this.  So, it would be great to

> > > have this fixed.

> > > 

> > > Philippe> There was a previous trial to fix this PR.

> > > Philippe> See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html

> > > Philippe> This trial was however only fixing the problem for the automatically

> > > Philippe> determined executable files when doing attach.

> > 

> > I have re-read all the exchanges for the above fix trial.

> > It ended up with the above msg00118 with Pedro asking 'WDYT?',

> > and it looks like everybody was thinking nothing.

> > So, after more than 4 years, here is some feedback :).

> > 

> > IMO, the distinction between auto-discovered and user-specified

> > is artificial: if the user will debug a PID with what looks like a

> > wrong executable specified by the user, GDB should not silently

> > use the user provided file, as keeping it is very probably

> > wrong (at least in the cases I have seen, it was always me

> > doing an error, and the patch below will make the error visible

> > and easy to fix) :

> > The patch proposed below will indicate that there is very probably

> > something weird, and then the user can change the setting and/or

> > decide to not load the exec file discovered from the PID.

> > 

> > Also, the patch above makes GDB unconditionally switch of exec file

> > for auto-discovered

> > files.

> > If ever the user really want to do something 'unusual', no way.

> > Again, the patch

> > below allows the user to control the behaviour

> > if really something unusual is

> > desired/needed.

> > 

> > > 

> > > There was also this series:

> > > 

> > > https://sourceware.org/ml/gdb-patches/2014-03/msg00476.html

> > > 

> > > ... which seemingly went through at least 4 rounds of review and then

> > > did not land for some reason.

> > 

> > After digging in the mail archive, it looks like the last state

> > of this mail thread was the v5 done by Jan Kratochvil.

> > I found no reaction to this v5.  No idea why it stalled.

> > 

> > That being said, I think that this last series does something different.

> > IIUC, this patch series ensures that when GDB receives the pathname

> > of a file to use (typically from a gdbserver), that GDB will use the

> > build-id to decide to *not* load the debug info, but this does not

> > lead to 'switch' to another correct exec-file.

> > 

> > In other words, this v5 ensures that GDB does not load wrong

> > debug info for a file GDB has decided to use as exec-file (or library).

> > It does not detect or report that a wrong file is selected for a PID.

> > 

> > In that sense, the patch proposed below is not contradictory with this

> > last patch v5: once GDB has decided to use a certain exec-file,

> > GDB today checks e.g. that the file has not changed of time stamp

> > (otherwise it re-reads the file).  As I understand, the build-id is just

> > a more sophisticated way to detect if the file loaded by the PID

> > is consistent with either the reported file to use by GDB and/or with

> > the separate debug info GDB finds on the host side.

> > 

> > > 

> > > Before approving anything I would like to understand why neither of

> > > these earlier attempts went in.

> > > 

> > > The build-id approach is appealing because (especially since the demise

> > > of prelink) it seems that gdb could simply trust it and abandon

> > > non-matching symbol files, without needing a "user supplied" bit.

> > 

> > In summary, I think that the patch below is another approach

> > than the 'sticky user executable' concept, but does not aim at

> > solving the problem of an exec-file or a remote exec-file

> > inconsistent with some GDB host debug info.

> > 

> > 

> > Philippe

> > 

> > 

> >
Pedro Alves April 10, 2019, 12:18 p.m. | #5
On 4/10/19 1:09 PM, Philippe Waroquiers wrote:
> Some feedback ?

> (or should I just go on, finish the work to submit an RFA?)

> Thanks

> Philippe

> 

> NB: Jan told me this was a long time ago, and could not give

> much comments, as he is not working on GDB anymore.

> 

> On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

>> Pedro/Jan,

>> Some feedback ?

>>


Sorry, it was on my list to reply back, but I wanted to dig in
to the archives first, because I don't really recall what happened
to the previous work.  Looks like there's a lot to digest, so
I'll need some time.

Thanks,
Pedro Alves
Philippe Waroquiers July 5, 2019, 7:49 p.m. | #6
Ping ?

Thanks

Philippe


On Wed, 2019-04-10 at 13:18 +0100, Pedro Alves wrote:
> On 4/10/19 1:09 PM, Philippe Waroquiers wrote:

> > Some feedback ?

> > (or should I just go on, finish the work to submit an RFA?)

> > Thanks

> > Philippe

> > 

> > NB: Jan told me this was a long time ago, and could not give

> > much comments, as he is not working on GDB anymore.

> > 

> > On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

> > > Pedro/Jan,

> > > Some feedback ?

> > > 

> 

> Sorry, it was on my list to reply back, but I wanted to dig in

> to the archives first, because I don't really recall what happened

> to the previous work.  Looks like there's a lot to digest, so

> I'll need some time.

> 

> Thanks,

> Pedro Alves
Philippe Waroquiers Aug. 7, 2019, 7:43 p.m. | #7
Thanks
Philippe

On Wed, 2019-04-10 at 13:18 +0100, Pedro Alves wrote:
> On 4/10/19 1:09 PM, Philippe Waroquiers wrote:

> > Some feedback ?

> > (or should I just go on, finish the work to submit an RFA?)

> > Thanks

> > Philippe

> > 

> > NB: Jan told me this was a long time ago, and could not give

> > much comments, as he is not working on GDB anymore.

> > 

> > On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

> > > Pedro/Jan,

> > > Some feedback ?

> > > 

> 

> Sorry, it was on my list to reply back, but I wanted to dig in

> to the archives first, because I don't really recall what happened

> to the previous work.  Looks like there's a lot to digest, so

> I'll need some time.

> 

> Thanks,

> Pedro Alves
Philippe Waroquiers Aug. 15, 2019, 1:21 p.m. | #8
Ping.
Thanks
Philippe



On Wed, 2019-08-07 at 21:43 +0200, Philippe Waroquiers wrote:
> Thanks

> Philippe

> 

> On Wed, 2019-04-10 at 13:18 +0100, Pedro Alves wrote:

> > On 4/10/19 1:09 PM, Philippe Waroquiers wrote:

> > > Some feedback ?

> > > (or should I just go on, finish the work to submit an RFA?)

> > > Thanks

> > > Philippe

> > > 

> > > NB: Jan told me this was a long time ago, and could not give

> > > much comments, as he is not working on GDB anymore.

> > > 

> > > On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

> > > > Pedro/Jan,

> > > > Some feedback ?

> > > > 

> > 

> > Sorry, it was on my list to reply back, but I wanted to dig in

> > to the archives first, because I don't really recall what happened

> > to the previous work.  Looks like there's a lot to digest, so

> > I'll need some time.

> > 

> > Thanks,

> > Pedro Alves
Philippe Waroquiers Sept. 7, 2019, 11:46 a.m. | #9
Ping.
Thanks
Philippe

On Wed, 2019-04-10 at 13:18 +0100, Pedro Alves wrote:
> On 4/10/19 1:09 PM, Philippe Waroquiers wrote:

> > Some feedback ?

> > (or should I just go on, finish the work to submit an RFA?)

> > Thanks

> > Philippe

> > 

> > NB: Jan told me this was a long time ago, and could not give

> > much comments, as he is not working on GDB anymore.

> > 

> > On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

> > > Pedro/Jan,

> > > Some feedback ?

> > > 

> 

> Sorry, it was on my list to reply back, but I wanted to dig in

> to the archives first, because I don't really recall what happened

> to the previous work.  Looks like there's a lot to digest, so

> I'll need some time.

> 

> Thanks,

> Pedro Alves
Philippe Waroquiers Sept. 29, 2019, 3:26 p.m. | #10
Some feedback ?

Thanks

Philippe

On Wed, 2019-04-10 at 13:18 +0100, Pedro Alves wrote:
> On 4/10/19 1:09 PM, Philippe Waroquiers wrote:

> > Some feedback ?

> > (or should I just go on, finish the work to submit an RFA?)

> > Thanks

> > Philippe

> > 

> > NB: Jan told me this was a long time ago, and could not give

> > much comments, as he is not working on GDB anymore.

> > 

> > On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

> > > Pedro/Jan,

> > > Some feedback ?

> > > 

> 

> Sorry, it was on my list to reply back, but I wanted to dig in

> to the archives first, because I don't really recall what happened

> to the previous work.  Looks like there's a lot to digest, so

> I'll need some time.

> 

> Thanks,

> Pedro Alves
Philippe Waroquiers Oct. 20, 2019, 10:24 a.m. | #11
Ping ?
(attaching a patch rebased on recent master)

Philippe

On Sun, 2019-09-29 at 17:26 +0200, Philippe Waroquiers wrote:
> Some feedback ?

> 

> Thanks

> 

> Philippe

> 

> On Wed, 2019-04-10 at 13:18 +0100, Pedro Alves wrote:

> > On 4/10/19 1:09 PM, Philippe Waroquiers wrote:

> > > Some feedback ?

> > > (or should I just go on, finish the work to submit an RFA?)

> > > Thanks

> > > Philippe

> > > 

> > > NB: Jan told me this was a long time ago, and could not give

> > > much comments, as he is not working on GDB anymore.

> > > 

> > > On Wed, 2019-04-03 at 22:24 +0200, Philippe Waroquiers wrote:

> > > > Pedro/Jan,

> > > > Some feedback ?

> > > > 

> > 

> > Sorry, it was on my list to reply back, but I wanted to dig in

> > to the archives first, because I don't really recall what happened

> > to the previous work.  Looks like there's a lot to digest, so

> > I'll need some time.

> > 

> > Thanks,

> > Pedro Alves
From fdd33c4d057723c35441a0840c9a84e76613a4af Mon Sep 17 00:00:00 2001
From: Philippe Waroquiers <philippe.waroquiers@skynet.be>
Date: Mon, 18 Mar 2019 23:57:32 +0100
Subject: [PATCH] Have an option to tell GDB to detect and possibly handle
 mismatched exec-files.

A recurrent problem with GDB is that GDB uses the wrong exec-file
when using the attach/detach commands sucessfully.
Also, in case the user specifies a file on the command line but attaches
to the wrong PID, this error is not made visible and gives a not user
understandable behaviour.

For example:
  $ gdb
  ...
  (gdb) atta 2682
  Attaching to process 2682
  [New LWP 2683]
  [New LWP 2684]
  [New LWP 2685]
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  0x00007f5ff829f603 in select () at ../sysdeps/unix/syscall-template.S:84
  84	../sysdeps/unix/syscall-template.S: No such file or directory.
  (gdb) det
  Detaching from program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 2682
  [Inferior 1 (process 2682) detached]
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  Reading symbols from /lib64/ld-linux-x86-64.so.2...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in ?? ()
  (gdb) bt
  #0  0x00007f43c23478a0 in ?? ()
  #1  0x0000558909e3ad91 in ?? ()
  #2  0x0000202962646700 in ?? ()
  #3  0x00007ffc69c74e70 in ?? ()
  #4  0x000055890c1d2350 in ?? ()
  #5  0x0000000000000000 in ?? ()
  (gdb)

The second attach has kept the executable of the first attach.
(in this case, 31069 is the PID of a GDB, that has nothing to do
with the first determined 'sleepers' executable).

Similarly, if specifying an executable, but attaching to a wrong pid,
we get:

  gdb /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  ...
  Reading symbols from /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers...
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  Reading symbols from /lib64/ld-linux-x86-64.so.2...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in ?? ()
  (gdb) bt
  #0  0x00007f43c23478a0 in ?? ()
  #1  0x0000558909e3ad91 in ?? ()
  #2  0x0000202962646700 in ?? ()
  #3  0x00007ffc69c74e70 in ?? ()
  #4  0x000055890c1d2350 in ?? ()
  #5  0x0000000000000000 in ?? ()
  (gdb)

And it is unclear to the user what has happened/what is going wrong.

This patch series implements a new option:
  (gdb) apropos exec-file-mismatch
  set exec-file-mismatch -- Set exec-file-mismatch handling
  show exec-file-mismatch -- Show exec-file-mismatch handling
  (gdb) help set exec-file-mismatch
  Set exec-file-mismatch handling.   (reload/warn/off)
  (gdb)

"reload" means: in case of mismatch between the current exec-file
and the automatically determined exec-file of the PID we are attaching to,
give a warning to the user and reload the automatically determined exec-file.
This reload might be subject to a question to the user.

"warn" means: in case of mismatch, just give a warning to the user.

"off" means: do not check for mismatch.

This fixes PR gdb/17626.
There was a previous trial to fix this PR.
See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html
This trial was however only fixing the problem for the automatically
determined executable files when doing attach.
It was differentiating the 'user specified executable files' ("sticky")
from the executable files automatically found by GDB.
But such user specified sticky executables are in most cases due
to a wrong manipulation by the user, giving unexpected results
such as backtrace showing no function like in the above example.

This patch ensures that whenever a process executable can be
determined, that the user is warned if there is a mismatch.

The same tests as above then give:

  (gdb) atta 2682
  Attaching to process 2682
  [New LWP 2683]
  [New LWP 2684]
  [New LWP 2685]
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  0x00007f5ff829f603 in select () at ../sysdeps/unix/syscall-template.S:84
  84	../sysdeps/unix/syscall-template.S: No such file or directory.
  (gdb) det
  Detaching from program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 2682
  [Inferior 1 (process 2682) detached]
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
  exec-file-mismatch handling is currently "reload"
  Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
  Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...
  Setting up the environment for debugging gdb.
  ...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in __poll_nocancel () at ../sysdeps/unix/syscall-template.S:84
  84	../sysdeps/unix/syscall-template.S: No such file or directory.
  (top-gdb) bt
  During symbol reading: incomplete CFI data; unspecified registers (e.g., rax) at 0x7f43c23478ad
  During symbol reading: unsupported tag: 'DW_TAG_unspecified_type'
  During symbol reading: cannot get low and high bounds for subprogram DIE at 0x12282a7
  During symbol reading: Child DIE 0x12288ba and its abstract origin 0x1228b26 have different parents
  During symbol reading: DW_AT_call_target target DIE has invalid low pc, for referencing DIE 0x1229540 [in module /bd/home/philippe/gdb/git/build_fixes/gdb/gdb]
  #0  0x00007f43c23478a0 in __poll_nocancel () at ../sysdeps/unix/syscall-template.S:84
  #1  0x0000558909e3ad91 in poll (__timeout=-1, __nfds=<optimized out>, __fds=<optimized out>) at /usr/include/x86_64-linux-gnu/bits/poll2.h:46
  #2  gdb_wait_for_event (block=block@entry=1) at ../../fixes/gdb/event-loop.c:772
  #3  0x0000558909e3aef4 in gdb_do_one_event () at ../../fixes/gdb/event-loop.c:347
  #4  0x0000558909e3b085 in gdb_do_one_event () at ../../fixes/gdb/common/common-exceptions.h:219
  #5  start_event_loop () at ../../fixes/gdb/event-loop.c:371
  During symbol reading: Member function "~_Sp_counted_base" (offset 0x1c69bf7) is virtual but the vtable offset is not specified
  During symbol reading: Multiple children of DIE 0x1c8f5a0 refer to DIE 0x1c8f0ee as their abstract origin
  #6  0x0000558909ed3b78 in captured_command_loop () at ../../fixes/gdb/main.c:331
  #7  0x0000558909ed4b6d in captured_main (data=<optimized out>) at ../../fixes/gdb/main.c:1174
  #8  gdb_main (args=<optimized out>) at ../../fixes/gdb/main.c:1190
  #9  0x0000558909c1e9a8 in main (argc=<optimized out>, argv=<optimized out>) at ../../fixes/gdb/gdb.c:32
  (top-gdb)

  gdb /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepersReading symbols from /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers...
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
  exec-file-mismatch handling is currently "reload"
  Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
  Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...
  Setting up the environment for debugging gdb.
  ....

In other words, it now works as intuitively expected by the user.
If ever the user gave the correct executable on the command line,
then attached to the wrong pid, then confirmed loading the wrong executable,
the user can simply fix this by detaching, and attaching to the correct pid,
GDB will then tell again to the user that the exec-file might better
be reloaded.

The default value of "reload" is chosen instead of e.g. "warn" as in most
cases, switching of executable will be the correct action,
and in any case, the user can decide to not load the executable.

Tested on debian/amd64, no regression.

Still missing:
  ChangeLog
  NEWS
  user manual
---
 gdb/exec.c    | 103 ++++++++++++++++++++++++++++++++++++++++++++++++++
 gdb/gdbcore.h |   5 +++
 gdb/infcmd.c  |   3 ++
 gdb/remote.c  |   3 ++
 4 files changed, 114 insertions(+)

diff --git a/gdb/exec.c b/gdb/exec.c
index 6bdf9abb3e..3ff45c920c 100644
--- a/gdb/exec.c
+++ b/gdb/exec.c
@@ -82,6 +82,49 @@ struct exec_target final : public target_ops
 
 static exec_target exec_ops;
 
+/* How to handle a mismatch between the current exec file and the exec
+   file determined from target.  */
+
+static const char *const exec_file_mismatch_names[]
+  = {"reload", "warn", "off", NULL };
+enum exec_file_mismatch_mode
+  {
+    exec_file_mismatch_reload, exec_file_mismatch_warn, exec_file_mismatch_off
+  };
+static const char *exec_file_mismatch = exec_file_mismatch_names[0];
+enum exec_file_mismatch_mode exec_file_mismatch_mode = exec_file_mismatch_reload;
+
+/* Show command.  */
+static void
+show_exec_file_mismatch_command (struct ui_file *file, int from_tty,
+				 struct cmd_list_element *c, const char *value)
+{
+  fprintf_filtered (gdb_stdout,
+		    _("exec-file-mismatch handling is currently \"%s\".\n"),
+		    exec_file_mismatch_names[exec_file_mismatch_mode]);
+}
+
+/* Set command.  Change the setting for range checking.  */
+static void
+set_exec_file_mismatch_command (const char *ignore,
+				int from_tty, struct cmd_list_element *c)
+{
+  for (enum exec_file_mismatch_mode mode = exec_file_mismatch_reload;
+       ;
+       mode = static_cast<enum exec_file_mismatch_mode>(1 + (int) mode))
+    {
+      if (strcmp (exec_file_mismatch, exec_file_mismatch_names[mode]) == 0)
+	{
+	  exec_file_mismatch_mode = mode;
+	  return;
+	}
+      if (mode == exec_file_mismatch_off)
+	internal_error (__FILE__, __LINE__,
+			_("Unrecognized exec-file-mismatch setting: \"%s\""),
+			exec_file_mismatch);
+    }
+}
+
 /* Whether to open exec and core files read-only or read-write.  */
 
 bool write_files = false;
@@ -191,6 +234,55 @@ try_open_exec_file (const char *exec_file_host, struct inferior *inf,
 
 /* See gdbcore.h.  */
 
+void
+validate_exec_file (int from_tty)
+{
+  if (exec_file_mismatch_mode == exec_file_mismatch_off)
+    return; /* User asked to do nothing.  */
+
+  char *current_exec_file = get_exec_file (0);
+  struct inferior *inf = current_inferior ();
+  /* Try to determine a filename from the process itself.  */
+  char *pid_exec_file = target_pid_to_exec_file (inf->pid);
+
+  if (current_exec_file == NULL || pid_exec_file == NULL)
+    return; /* We cannot validate the exec file.  */
+
+  std::string exec_file_target (pid_exec_file);
+
+  /* In case the exec file is not local, exec_file_target has to point at
+     the target file system.  */
+  if (is_target_filename (current_exec_file) && !target_filesystem_is_local ())
+    exec_file_target = TARGET_SYSROOT_PREFIX + exec_file_target;
+
+  if (strcmp (exec_file_target.c_str (), current_exec_file) != 0)
+    {
+      warning (_("Mismatch between current exec-file %s\n"
+		 "and automatically determined exec-file %s\n"
+		 "exec-file-mismatch handling is currently \"%s\""),
+	       current_exec_file, exec_file_target.c_str (),
+	       exec_file_mismatch_names[exec_file_mismatch_mode]);
+      if (exec_file_mismatch_mode == exec_file_mismatch_reload)
+	{
+	  symfile_add_flags add_flags = SYMFILE_MAINLINE;
+	  if (from_tty)
+	    add_flags |= SYMFILE_VERBOSE;
+	  try
+	    {
+	      symbol_file_add_main (exec_file_target.c_str (), add_flags);
+	      exec_file_attach (exec_file_target.c_str (), from_tty);
+	    }
+	  catch (gdb_exception_error &err)
+	    {
+	      warning (_("reloading %s %s"), exec_file_target.c_str (),
+		       err.message != NULL ? err.what () : "error");
+	    }
+	}
+    }
+}
+
+/* See gdbcore.h.  */
+
 void
 exec_file_locate_attach (int pid, int defer_bp_reset, int from_tty)
 {
@@ -1070,5 +1162,16 @@ Show writing into executable and core files."), NULL,
 			   show_write_files,
 			   &setlist, &showlist);
 
+  add_setshow_enum_cmd ("exec-file-mismatch", class_support,
+			exec_file_mismatch_names,
+			&exec_file_mismatch,
+			_("\
+Set exec-file-mismatch handling.   (reload/warn/off)"),
+			_("\
+Show exec-file-mismatch handling.  (reload/warn/off"),
+			NULL, set_exec_file_mismatch_command,
+			show_exec_file_mismatch_command,
+			&setlist, &showlist);
+
   add_target (exec_target_info, exec_target_open, filename_completer);
 }
diff --git a/gdb/gdbcore.h b/gdb/gdbcore.h
index cac91c097b..0f3c6aad95 100644
--- a/gdb/gdbcore.h
+++ b/gdb/gdbcore.h
@@ -157,6 +157,11 @@ extern void exec_file_locate_attach (int pid, int defer_bp_reset, int from_tty);
 
 extern void validate_files (void);
 
+/* Produce a warning if the current exec file does not match the exec
+   file determined from the target.  In case of mismatch, ask the user
+   if the exec file determined from target must be loaded.  */
+extern void validate_exec_file (int from_tty);
+
 /* The current default bfd target.  */
 
 extern char *gnutarget;
diff --git a/gdb/infcmd.c b/gdb/infcmd.c
index 465d3a10e5..4ca092c44d 100644
--- a/gdb/infcmd.c
+++ b/gdb/infcmd.c
@@ -2563,6 +2563,9 @@ setup_inferior (int from_tty)
       reread_symbols ();
     }
 
+  /* Check for exec file mismatch, and let the user solve it.  */
+  validate_exec_file (from_tty);
+
   /* Take any necessary post-attaching actions for this platform.  */
   target_post_attach (inferior_ptid.pid ());
 
diff --git a/gdb/remote.c b/gdb/remote.c
index 5e1745db44..98346b321f 100644
--- a/gdb/remote.c
+++ b/gdb/remote.c
@@ -2380,6 +2380,9 @@ remote_target::remote_add_inferior (bool fake_pid_p, int pid, int attached,
   if (try_open_exec && get_exec_file (0) == NULL)
     exec_file_locate_attach (pid, 0, 1);
 
+  /* Check for exec file mismatch, and let the user solve it.  */
+  validate_exec_file (1);
+
   return inf;
 }
Tom Tromey Dec. 13, 2019, 10:09 p.m. | #12
>>>>> "Philippe" == Philippe Waroquiers <philippe.waroquiers@skynet.be> writes:


Philippe> Ping ?
Philippe> (attaching a patch rebased on recent master)

I'm sorry for the extreme delay on this.

I read through the thread a bit, and I looked briefly at the other
threads in the past.  I didn't read those exhaustively though.

Philippe> This patch series implements a new option:
Philippe>   (gdb) apropos exec-file-mismatch
Philippe>   set exec-file-mismatch -- Set exec-file-mismatch handling
Philippe>   show exec-file-mismatch -- Show exec-file-mismatch handling

Seems reasonable.

Philippe> The same tests as above then give:
[...]

Philippe>   (gdb) atta 31069
Philippe>   Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
Philippe>   warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
Philippe>   and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
Philippe>   exec-file-mismatch handling is currently "reload"
Philippe>   Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
Philippe>   Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...

Nice.

Philippe> +enum exec_file_mismatch_mode exec_file_mismatch_mode = exec_file_mismatch_reload;

This can be "static".

Philippe> Still missing:
Philippe>   ChangeLog
Philippe>   NEWS
Philippe>   user manual

... still needs these, but I completely understand why you put off
writing them.

Philippe> +void
Philippe> +validate_exec_file (int from_tty)
Philippe> +{
Philippe> +  if (exec_file_mismatch_mode == exec_file_mismatch_off)
Philippe> +    return; /* User asked to do nothing.  */

Normally gdb prefers comments not to be at the end of the line like
this.  It's fine to just stick them before the "if", maybe reworded
sometimes.

Philippe> +  char *current_exec_file = get_exec_file (0);

Could be const.

Philippe> +  /* Try to determine a filename from the process itself.  */
Philippe> +  char *pid_exec_file = target_pid_to_exec_file (inf->pid);

Also could be const.

Philippe> +  if (strcmp (exec_file_target.c_str (), current_exec_file) != 0)

This can just use !=, because exec_file_target is a std::string.

Philippe> +    {
Philippe> +      warning (_("Mismatch between current exec-file %s\n"
Philippe> +		 "and automatically determined exec-file %s\n"
Philippe> +		 "exec-file-mismatch handling is currently \"%s\""),

I wonder whether the newlines are needed?  And also if now this should
use %ps and styled_string and file_name_style?


What happens if the inferior exec()s something else?  Also why was a
validate call required in remote.c but not in any other target?

thanks,
Tom

Patch

diff --git a/gdb/exec.c b/gdb/exec.c
index 77bd140a8e..054753d793 100644
--- a/gdb/exec.c
+++ b/gdb/exec.c
@@ -82,6 +82,49 @@  struct exec_target final : public target_ops
 
 static exec_target exec_ops;
 
+/* How to handle a mismatch between the current exec file and the exec
+   file determined from target.  */
+
+static const char *const exec_file_mismatch_names[]
+  = {"reload", "warn", "off", NULL };
+enum exec_file_mismatch_mode
+  {
+    exec_file_mismatch_reload, exec_file_mismatch_warn, exec_file_mismatch_off
+  };
+static const char *exec_file_mismatch = exec_file_mismatch_names[0];
+enum exec_file_mismatch_mode exec_file_mismatch_mode = exec_file_mismatch_reload;
+
+/* Show command.  */
+static void
+show_exec_file_mismatch_command (struct ui_file *file, int from_tty,
+				 struct cmd_list_element *c, const char *value)
+{
+  fprintf_filtered (gdb_stdout,
+		    _("exec-file-mismatch handling is currently \"%s\".\n"),
+		    exec_file_mismatch_names[exec_file_mismatch_mode]);
+}
+
+/* Set command.  Change the setting for range checking.  */
+static void
+set_exec_file_mismatch_command (const char *ignore,
+				int from_tty, struct cmd_list_element *c)
+{
+  for (enum exec_file_mismatch_mode mode = exec_file_mismatch_reload;
+       ;
+       mode = static_cast<enum exec_file_mismatch_mode>(1 + (int) mode))
+    {
+      if (strcmp (exec_file_mismatch, exec_file_mismatch_names[mode]) == 0)
+	{
+	  exec_file_mismatch_mode = mode;
+	  return;
+	}
+      if (mode == exec_file_mismatch_off)
+	internal_error (__FILE__, __LINE__,
+			_("Unrecognized exec-file-mismatch setting: \"%s\""),
+			exec_file_mismatch);
+    }
+}
+
 /* Whether to open exec and core files read-only or read-write.  */
 
 int write_files = 0;
@@ -201,6 +244,56 @@  try_open_exec_file (const char *exec_file_host, struct inferior *inf,
 
 /* See gdbcore.h.  */
 
+void
+validate_exec_file (int from_tty)
+{
+  if (exec_file_mismatch_mode == exec_file_mismatch_off)
+    return; /* User asked to do nothing.  */
+
+  char *current_exec_file = get_exec_file (0);
+  struct inferior *inf = current_inferior ();
+  /* Try to determine a filename from the process itself.  */
+  char *pid_exec_file = target_pid_to_exec_file (inf->pid);
+
+  if (current_exec_file == NULL || pid_exec_file == NULL)
+    return; /* We cannot validate the exec file.  */
+
+  std::string exec_file_target (pid_exec_file);
+
+  /* In case the exec file is not local, exec_file_target has to point at
+     the target file system.  */
+  if (is_target_filename (current_exec_file) && !target_filesystem_is_local ())
+    exec_file_target = TARGET_SYSROOT_PREFIX + exec_file_target;
+
+  if (strcmp (exec_file_target.c_str (), current_exec_file) != 0)
+    {
+      warning (_("Mismatch between current exec-file %s\n"
+		 "and automatically determined exec-file %s\n"
+		 "exec-file-mismatch handling is currently \"%s\""),
+	       current_exec_file, exec_file_target.c_str (),
+	       exec_file_mismatch_names[exec_file_mismatch_mode]);
+      if (exec_file_mismatch_mode == exec_file_mismatch_reload)
+	{
+	  symfile_add_flags add_flags = SYMFILE_MAINLINE;
+	  if (from_tty)
+	    add_flags |= SYMFILE_VERBOSE;
+	  TRY
+	    {
+	      symbol_file_add_main (exec_file_target.c_str (), add_flags);
+	      exec_file_attach (exec_file_target.c_str (), from_tty);
+	    }
+	  CATCH (err, RETURN_MASK_ERROR)
+	    {
+	      warning (_("reloading %s %s"), exec_file_target.c_str (),
+		       err.message != NULL ? err.message : "error");
+	    }
+	  END_CATCH
+	}
+    }
+}
+
+/* See gdbcore.h.  */
+
 void
 exec_file_locate_attach (int pid, int defer_bp_reset, int from_tty)
 {
@@ -1084,5 +1177,16 @@  Show writing into executable and core files."), NULL,
 			   show_write_files,
 			   &setlist, &showlist);
 
+  add_setshow_enum_cmd ("exec-file-mismatch", class_support,
+			exec_file_mismatch_names,
+			&exec_file_mismatch,
+			_("\
+Set exec-file-mismatch handling.   (reload/warn/off)"),
+			_("\
+Show exec-file-mismatch handling.  (reload/warn/off"),
+			NULL, set_exec_file_mismatch_command,
+			show_exec_file_mismatch_command,
+			&setlist, &showlist);
+
   add_target (exec_target_info, exec_target_open, filename_completer);
 }
diff --git a/gdb/gdbcore.h b/gdb/gdbcore.h
index 6f9b4d88ec..611a389e22 100644
--- a/gdb/gdbcore.h
+++ b/gdb/gdbcore.h
@@ -157,6 +157,11 @@  extern void exec_file_locate_attach (int pid, int defer_bp_reset, int from_tty);
 
 extern void validate_files (void);
 
+/* Produce a warning if the current exec file does not match the exec
+   file determined from the target.  In case of mismatch, ask the user
+   if the exec file determined from target must be loaded.  */
+extern void validate_exec_file (int from_tty);
+
 /* The current default bfd target.  */
 
 extern char *gnutarget;
diff --git a/gdb/infcmd.c b/gdb/infcmd.c
index c5977c48a9..5812827ae5 100644
--- a/gdb/infcmd.c
+++ b/gdb/infcmd.c
@@ -2560,6 +2560,9 @@  setup_inferior (int from_tty)
       reread_symbols ();
     }
 
+  /* Check for exec file mismatch, and let the user solve it.  */
+  validate_exec_file (from_tty);
+
   /* Take any necessary post-attaching actions for this platform.  */
   target_post_attach (inferior_ptid.pid ());
 
diff --git a/gdb/remote.c b/gdb/remote.c
index 657a4a25ca..c0c80ee140 100644
--- a/gdb/remote.c
+++ b/gdb/remote.c
@@ -2392,6 +2392,9 @@  remote_target::remote_add_inferior (int fake_pid_p, int pid, int attached,
   if (try_open_exec && get_exec_file (0) == NULL)
     exec_file_locate_attach (pid, 0, 1);
 
+  /* Check for exec file mismatch, and let the user solve it.  */
+  validate_exec_file (1);
+
   return inf;
 }