restart for remote target

restart for remote target

[Efim Monyak]

Hi all,

I need an restart command for remote target. Set Program Counter to i.e. 
0 is not an
option because interrupts.
I implemented it as user-defined command but all breakpoints must be 
delete before load.
After load all breakpoints are lost.
Is it possible to save breakpoint data in order restore it after 
application reload?

Thanks

Re: restart for remote target

[Daniel Jacobowitz]

On Tue, Jul 25, 2006 at 05:42:14PM +0200, Efim Monyak wrote:
> Hi all,
> 
> I need an restart command for remote target. Set Program Counter to i.e. 
> 0 is not an
> option because interrupts.
> I implemented it as user-defined command but all breakpoints must be 
> delete before load.
> After load all breakpoints are lost.
> Is it possible to save breakpoint data in order restore it after 
> application reload?

Sorry, but your explanation omits too much; we need more information.
Load should not remove breakpoints.

Would it help you to connect using 'target extended-remote', and then
use "run" to restart the target?  That sends an 'R' packet to the stub,
which can do whatever it needs to.

-- 
Daniel Jacobowitz
CodeSourcery

Re: restart for remote target

[Efim Monyak]

Daniel Jacobowitz schrieb:
> On Tue, Jul 25, 2006 at 05:42:14PM +0200, Efim Monyak wrote:
>   
>> Hi all,
>>
>> I need an restart command for remote target. Set Program Counter to i.e. 
>> 0 is not an
>> option because interrupts.
>> I implemented it as user-defined command but all breakpoints must be 
>> delete before load.
>> After load all breakpoints are lost.
>> Is it possible to save breakpoint data in order restore it after 
>> application reload?
>>     
>
> Sorry, but your explanation omits too much; we need more information.
> Load should not remove breakpoints.
>
> Would it help you to connect using 'target extended-remote', and then
> use "run" to restart the target?  That sends an 'R' packet to the stub,
> which can do whatever it needs to.
>
>   
You are right.
I delete all breakpoints before load the applications again because my 
Hardware supports only two
breakpoints. In other case I receive error from GDB. The load is 
multistep. In one of steps I must set
an breakpoint to particular address and only one way I found to put it 
away is command delete all breakpoints.
As I see no command deletes breakpoint on address. Use breakpoint to 
source line is not very good on this place.

It is not possible for stub restart the application. They are often 
problem if a remote target is used.
As I wrote set PC to start address works not properly if i.e. interrupts 
are enabled.
If application runs from RAM it is lost after Hardware reset.

The only handy way I see is to reload application set all used 
breakpoints and start it again.

Re: restart for remote target

[Daniel Jacobowitz]

On Tue, Jul 25, 2006 at 06:19:51PM +0200, Efim Monyak wrote:
> I delete all breakpoints before load the applications again because my 
> Hardware supports only two
> breakpoints. In other case I receive error from GDB. The load is 
> multistep. In one of steps I must set
> an breakpoint to particular address and only one way I found to put it 
> away is command delete all breakpoints.
> As I see no command deletes breakpoint on address. Use breakpoint to 
> source line is not very good on this place.

"delete *0x3000000" should work.

Also, it sounds like you want "disable", not "delete".

> It is not possible for stub restart the application. They are often 
> problem if a remote target is used.
> As I wrote set PC to start address works not properly if i.e. interrupts 
> are enabled.
> If application runs from RAM it is lost after Hardware reset.
> 
> The only handy way I see is to reload application set all used 
> breakpoints and start it again.

In many cases the stub can handle all of this automatically,
manually disabling interrupts et cetera.

-- 
Daniel Jacobowitz
CodeSourcery

Re: restart for remote target

[Efim Monyak]

Daniel Jacobowitz schrieb:
> On Tue, Jul 25, 2006 at 06:19:51PM +0200, Efim Monyak wrote:
>   
>> I delete all breakpoints before load the applications again because my 
>> Hardware supports only two
>> breakpoints. In other case I receive error from GDB. The load is 
>> multistep. In one of steps I must set
>> an breakpoint to particular address and only one way I found to put it 
>> away is command delete all breakpoints.
>> As I see no command deletes breakpoint on address. Use breakpoint to 
>> source line is not very good on this place.
>>     
>
> "delete *0x3000000" should work.
>
> Also, it sounds like you want "disable", not "delete".
>
>   
>> It is not possible for stub restart the application. They are often 
>> problem if a remote target is used.
>> As I wrote set PC to start address works not properly if i.e. interrupts 
>> are enabled.
>> If application runs from RAM it is lost after Hardware reset.
>>
>> The only handy way I see is to reload application set all used 
>> breakpoints and start it again.
>>     
>
> In many cases the stub can handle all of this automatically,
> manually disabling interrupts et cetera.
>
>   

debug session breaks because error.

hier is my user-defined command:

set $myvar = 0

define run
    if $myvar
        printf "\n 000000000000000000000000000 \n"
        disable
       
        printf "\n 44444444444444444444444444444 \n"
       
        k
        set remotebaud 300
        target remote /dev/ttyS3
        load reset.elf
        symbol-file reset.elf
        # set breakpoint to reset vector
        b * 0

        printf "\n 555555555555555555555555555555 \n"

        # start program
        c
        delete * 0
       
        printf "\n 666666666666666666666666666666 \n"
       
        # kill program
        k

        printf "\n 222222222222222222222222222 \n"

        set remotebaud 300
        target remote /dev/ttyS3
        load init.elf
        symbol-file init.elf
        # set breakpoint to end of program
        b end_init
        # start program
        c
        # delete all breakpoints (werden nicht mehr benutzt)
        clear end_init
        #kill program
        k
       
        printf "\n 33333333333333333333333333333 \n"
       
        set remotebaud 300
        target remote /dev/ttyS3
        load sample.elf
        symbol-file sample.elf
        set $myvar = 1
        enable
        continue
    else
        set $myvar = 1
        printf "\n 111111111111111111111111111 \n"
        continue
    end


and hier GDB messages:
156-exec-run
 (gdb)
156^running
 (gdb)
 ~"\n 000000000000000000000000000 \n"
 ~"\n 44444444444444444444444444444 \n"
156+download,{section="startup",section-size="32",total-size
="19040"}
156+download,{section="startup1",section-size="264",total-si
ze="19040"}
156+download,{section="prog",section-size="100",total-size="
19040"}
&"Error in re-setting breakpoint 3:\n"
&"No line 120 in file \"../src/main.c\".\n"
 ~"\n 555555555555555555555555555555 \n"
&"warning: bad breakpoint number at or near '* 0'\n"
&"warning: bad breakpoint number at or near '0'\n"
~"\n 666666666666666666666666666666 \n"
~"\n 222222222222222222222222222 \n"
156+download,{section="startup",section-size="32",total-size
="25297"}
156+download,{section="prog",section-size="4900",total-size=
"25297"}
156+download,{section="prog",section-sent="4051",section-siz
e="4900",total-sent=\
"4083",total-size="25297"}
~"\n 33333333333333333333333333333 \n"
156+download,{section="startup",section-size="516",total-siz
e="1132725"}
156+download,{section="text",section-size="1916",total-size=
"1132725"}
156+download,{section=".data",section-size="2",total-size="1
132725"}
&"Warning:\n"
&"Cannot insert breakpoint 6.\n"
&"Error accessing memory address 0x0: Unknown error 42949672
95.\n"
&"\n"
156^error,msg="Warning:\nCannot insert breakpoint 6.\nError
accessing memory add\
ress 0x0: Unknown error 4294967295.\n"


-- 
Mit freundlichen Gruessen
Efim Monjak


Lipowsky Industrie-Elektronik GmbH
64291 Darmstadt, Roemerstr. 57
Telefon:  +49-(0)6151-93591-0   
Telefax:  +49-(0)6151-93591-28  
Email:    ymonyak@lipowsky.de
Homepage: http://www.lipowsky.de
DIN EN ISO 9001:2000 certified by DQS

Re: restart for remote target

[Daniel Jacobowitz]

On Tue, Jul 25, 2006 at 06:54:09PM +0200, Efim Monyak wrote:
>        symbol-file reset.elf

If you want to use symbol files, you're out of luck.  When you change
the symbol file you're going to mess up all your breakpoints.

You might be able to get things to work by not changing the symbol
file.

-- 
Daniel Jacobowitz
CodeSourcery

Re: restart for remote target

[Efim Monyak]

Daniel Jacobowitz schrieb:
> On Tue, Jul 25, 2006 at 06:54:09PM +0200, Efim Monyak wrote:
>   
>>        symbol-file reset.elf
>>     
>
> If you want to use symbol files, you're out of luck.  When you change
> the symbol file you're going to mess up all your breakpoints.
>
> You might be able to get things to work by not changing the symbol
> file.
>
>   
Thanks, without line "symbol-file reset.elf" I have no more error
&"Error in re-setting breakpoint 3:\n"
&"No line 120 in file \"../src/main.c\".\n"

but still
156^error,msg="Warning:\nCannot insert breakpoint 6.\nError
accessing memory add\
ress 0x0: Unknown error 4294967295.\n"


Only after I use "tbreak * 0" ensteed "b * 0" and "delete * 0" it works.
Possibly delete * 0 don't working for GDB with ARM remote target.
I see Z0,0 command to stub after this breakpoint must be deleted
by delete * 0, it is a third breakpoint and it is an error for stub because
only two breakpoints are allowed.

The lines "symbol-file init.elf"   and "b end_init" are not changed and
they make no problem. Is the error depend for particular breakpoint type?
I mean breakpoint to address is handled in other way as i.e. breakpoint to
function.

Re: restart for remote target

[Daniel Jacobowitz]

On Wed, Jul 26, 2006 at 09:37:40AM +0200, Efim Monyak wrote:
> Only after I use "tbreak * 0" ensteed "b * 0" and "delete * 0" it works.
> Possibly delete * 0 don't working for GDB with ARM remote target.
> I see Z0,0 command to stub after this breakpoint must be deleted
> by delete * 0, it is a third breakpoint and it is an error for stub because
> only two breakpoints are allowed.
> 
> The lines "symbol-file init.elf"   and "b end_init" are not changed and
> they make no problem. Is the error depend for particular breakpoint type?
> I mean breakpoint to address is handled in other way as i.e. breakpoint to
> function.

No.

I have no idea what that error might be, sorry.

-- 
Daniel Jacobowitz
CodeSourcery

Re: restart for remote target

[Efim Monyak]

Daniel Jacobowitz schrieb:
> On Tue, Jul 25, 2006 at 06:19:51PM +0200, Efim Monyak wrote:
>   
>> I delete all breakpoints before load the applications again because my 
>> Hardware supports only two
>> breakpoints. In other case I receive error from GDB. The load is 
>> multistep. In one of steps I must set
>> an breakpoint to particular address and only one way I found to put it 
>> away is command delete all breakpoints.
>> As I see no command deletes breakpoint on address. Use breakpoint to 
>> source line is not very good on this place.
>>     
>
> "delete *0x3000000" should work.
>
> Also, it sounds like you want "disable", not "delete".
>
>   

Yes, is where a possibility to save breakpoint state (number, enable or 
disable) before disable
breakpoints and apply it after enable?
Because after "disable" command the information about breakpoint state 
is lost. They all are disabled.
I must do it by GDB script.
Or is it possible to define an array as long as maximal possible number 
of breakpoints?
I don't found a maximal possible value for $bpnum.

>> It is not possible for stub restart the application. They are often 
>> problem if a remote target is used.
>> As I wrote set PC to start address works not properly if i.e. interrupts 
>> are enabled.
>> If application runs from RAM it is lost after Hardware reset.
>>
>> The only handy way I see is to reload application set all used 
>> breakpoints and start it again.
>>     
>
> In many cases the stub can handle all of this automatically,
> manually disabling interrupts et cetera.
>
>   

stub receives command "insert breakpoint" for all enabled breakpoints 
before program step or continue and
"remove breakpoint" after program stop.

stub don't know breakpoints to be used on the next program step or continue.

Re: restart for remote target

[Daniel Jacobowitz]

On Thu, Jul 27, 2006 at 01:37:05PM +0200, Efim Monyak wrote:
> Yes, is where a possibility to save breakpoint state (number, enable or 
> disable) before disable
> breakpoints and apply it after enable?

No.  None of these things you want to do are possible in the GDB
scripting language; it just isn't complex enough.  Sorry.  You could
possibly drive GDB using something external and the machine interface
(MI).

There are tentative plans to hook up a real scripting language, but
they aren't ready yet - it may be a while.

> Because after "disable" command the information about breakpoint state 
> is lost. They all are disabled.
> I must do it by GDB script.
> Or is it possible to define an array as long as maximal possible number 
> of breakpoints?
> I don't found a maximal possible value for $bpnum.

There isn't a maximum.

> >In many cases the stub can handle all of this automatically,
> >manually disabling interrupts et cetera.
> >
> >  
> 
> stub receives command "insert breakpoint" for all enabled breakpoints 
> before program step or continue and
> "remove breakpoint" after program stop.
> 
> stub don't know breakpoints to be used on the next program step or continue.

I meant have the stub completely handle the restart operation, loading,
executing the bootloader, et cetera.

-- 
Daniel Jacobowitz
CodeSourcery