[ECOS] Problems when aborting tftp transfers

[ECOS] Problems when aborting tftp transfers

[Øyvind Harboe]

I'm seing problems when aborting tftp transfers:

- With CYGSEM_NET_TFTPD_MULTITHREADED == 0, a subsequent tftp transfer
after an aborted tftp transfer can get "stuck" since the main tftp
server loop is receiving stale packets and hence e.g. _write_file() is
invoked for the terminated tftp transfer session(there can be as many as
a dozen outstanding stale packets, hence tftp can take a *really* long
time to return to normal operation). Q: Is
CYGSEM_NET_TFTPD_MULTITHREADED == 0 working as intended for aborted tftp
transfers?

- I'm not sure this is a problem or just the way tftp works. When I set
CYGSEM_NET_TFTPD_MULTITHREADED==1, tftp works even if I abort transfers,
but the current tftp transfer session on the eCos end does not return to
the main tftp server loop until the current transfer session has
exhausted its retries. 


-- 
Øyvind Harboe
http://www.zylin.com



--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Andrew Lunn]

On Wed, Jul 21, 2004 at 12:47:53PM +0200, ?yvind Harboe wrote:
> I'm seing problems when aborting tftp transfers:
> 
> - With CYGSEM_NET_TFTPD_MULTITHREADED == 0, a subsequent tftp transfer
> after an aborted tftp transfer can get "stuck" since the main tftp
> server loop is receiving stale packets and hence e.g. _write_file() is
> invoked for the terminated tftp transfer session(there can be as many as
> a dozen outstanding stale packets, hence tftp can take a *really* long
> time to return to normal operation). Q: Is
> CYGSEM_NET_TFTPD_MULTITHREADED == 0 working as intended for aborted tftp
> transfers?
> 
> - I'm not sure this is a problem or just the way tftp works. When I set
> CYGSEM_NET_TFTPD_MULTITHREADED==1, tftp works even if I abort transfers,
> but the current tftp transfer session on the eCos end does not return to
> the main tftp server loop until the current transfer session has
> exhausted its retries. 

I don't follow what you are taking about here. Could you give some
more details, describe how you produce the problems, what goes wrong
and how you would expect it to act.

        Thanks
                Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Øyvind Harboe]

On Mon, 2004-08-02 at 15:01, Andrew Lunn wrote:
> On Wed, Jul 21, 2004 at 12:47:53PM +0200, ?yvind Harboe wrote:
> > I'm seing problems when aborting tftp transfers:
> > 
> > - With CYGSEM_NET_TFTPD_MULTITHREADED == 0, a subsequent tftp transfer
> > after an aborted tftp transfer can get "stuck" since the main tftp
> > server loop is receiving stale packets and hence e.g. _write_file() is
> > invoked for the terminated tftp transfer session(there can be as many as
> > a dozen outstanding stale packets, hence tftp can take a *really* long
> > time to return to normal operation). Q: Is
> > CYGSEM_NET_TFTPD_MULTITHREADED == 0 working as intended for aborted tftp
> > transfers?
> > 
> > - I'm not sure this is a problem or just the way tftp works. When I set
> > CYGSEM_NET_TFTPD_MULTITHREADED==1, tftp works even if I abort transfers,
> > but the current tftp transfer session on the eCos end does not return to
> > the main tftp server loop until the current transfer session has
> > exhausted its retries. 
> 
> I don't follow what you are taking about here. Could you give some
> more details, describe how you produce the problems, what goes wrong
> and how you would expect it to act.

here is what I did:

- with CYGSEM_NET_TFTPD_MULTITHREADED=0.
- started a tftp transfer from Windows machine
- aborted it on the Windows end
- at this point the tftp server goes back to the main loop and sees
multiple tftp transfer initiation requests from the aborted transfer
session. Eventually, the tftp server times out and recovers, but his
takes a *very* long time(minutes at least).
- since I only have a single tftp server thread, subsequent tftp
transfer requests will fail while the tftp server is busy timing out on
the obsolete session initiation requests.

If this rewording does not help explain things, I could try formulating
a patch.

>         Thanks
>                 Andrew
-- 
Øyvind Harboe
http://www.zylin.com



--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Andrew Lunn]

On Mon, Aug 02, 2004 at 03:45:01PM +0200, ?yvind Harboe wrote:
> On Mon, 2004-08-02 at 15:01, Andrew Lunn wrote:
> > On Wed, Jul 21, 2004 at 12:47:53PM +0200, ?yvind Harboe wrote:
> > > I'm seing problems when aborting tftp transfers:
> > > 
> > > - With CYGSEM_NET_TFTPD_MULTITHREADED == 0, a subsequent tftp transfer
> > > after an aborted tftp transfer can get "stuck" since the main tftp
> > > server loop is receiving stale packets and hence e.g. _write_file() is
> > > invoked for the terminated tftp transfer session(there can be as many as
> > > a dozen outstanding stale packets, hence tftp can take a *really* long
> > > time to return to normal operation). Q: Is
> > > CYGSEM_NET_TFTPD_MULTITHREADED == 0 working as intended for aborted tftp
> > > transfers?
> > > 
> > > - I'm not sure this is a problem or just the way tftp works. When I set
> > > CYGSEM_NET_TFTPD_MULTITHREADED==1, tftp works even if I abort transfers,
> > > but the current tftp transfer session on the eCos end does not return to
> > > the main tftp server loop until the current transfer session has
> > > exhausted its retries. 
> > 
> > I don't follow what you are taking about here. Could you give some
> > more details, describe how you produce the problems, what goes wrong
> > and how you would expect it to act.
> 
> here is what I did:
> 
> - with CYGSEM_NET_TFTPD_MULTITHREADED=0.
> - started a tftp transfer from Windows machine
> - aborted it on the Windows end

Does the M$ end send an error packet to terminate the transfer or
simply stop sending data packets?

> - at this point the tftp server goes back to the main loop and sees
> multiple tftp transfer initiation requests from the aborted transfer
> session. 

Initiation requests? You mean WRQ packets from the tftp client you
have just killed? That would be strange.
        
        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Øyvind Harboe]

On Mon, 2004-08-02 at 16:02, Andrew Lunn wrote:
> On Mon, Aug 02, 2004 at 03:45:01PM +0200, ?yvind Harboe wrote:
> > On Mon, 2004-08-02 at 15:01, Andrew Lunn wrote:
> > > On Wed, Jul 21, 2004 at 12:47:53PM +0200, ?yvind Harboe wrote:
> > > > I'm seing problems when aborting tftp transfers:
> > > > 
> > > > - With CYGSEM_NET_TFTPD_MULTITHREADED == 0, a subsequent tftp transfer
> > > > after an aborted tftp transfer can get "stuck" since the main tftp
> > > > server loop is receiving stale packets and hence e.g. _write_file() is
> > > > invoked for the terminated tftp transfer session(there can be as many as
> > > > a dozen outstanding stale packets, hence tftp can take a *really* long
> > > > time to return to normal operation). Q: Is
> > > > CYGSEM_NET_TFTPD_MULTITHREADED == 0 working as intended for aborted tftp
> > > > transfers?
> > > > 
> > > > - I'm not sure this is a problem or just the way tftp works. When I set
> > > > CYGSEM_NET_TFTPD_MULTITHREADED==1, tftp works even if I abort transfers,
> > > > but the current tftp transfer session on the eCos end does not return to
> > > > the main tftp server loop until the current transfer session has
> > > > exhausted its retries. 
> > > 
> > > I don't follow what you are taking about here. Could you give some
> > > more details, describe how you produce the problems, what goes wrong
> > > and how you would expect it to act.
> > 
> > here is what I did:
> > 
> > - with CYGSEM_NET_TFTPD_MULTITHREADED=0.
> > - started a tftp transfer from Windows machine
> > - aborted it on the Windows end
> 
> Does the M$ end send an error packet to terminate the transfer or
> simply stop sending data packets?

I don't know what ctrl-c does(I haven't got the equipment wired up so I
can't test right now), but surely terminating the process would cause an
error packet not to be transferred?


> > - at this point the tftp server goes back to the main loop and sees
> > multiple tftp transfer initiation requests from the aborted transfer
> > session. 
> 
> Initiation requests? You mean WRQ packets from the tftp client you
> have just killed? 

Yes.

I really don't speak PPP, TCP/IP, Windows and tftp well enough to pick
this apart without more study.

Perhaps this is something that was lingering in the PPP channel or
tcp/ip stack... 

> That would be strange.

Setting CYGSEM_NET_TFTPD_MULTITHREADED=1 fixes the problem because the
original port is closed + reopened.

>         
>         Andrew
-- 
Øyvind Harboe
http://www.zylin.com



--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Andrew Lunn]

> > Does the M$ end send an error packet to terminate the transfer or
> > simply stop sending data packets?
> 
> I don't know what ctrl-c does(I haven't got the equipment wired up so I
> can't test right now), but surely terminating the process would cause an
> error packet not to be transferred?

In the unix world, i would expect the application to trap ^C so that
it can do a clean shutdown. I don't know if M$ does this. Anyway, this
is probably not the real problem, its only an optimization. Since TFTP
is based on UDP, the error packet can get lost so the real method is
based on timeouts. It should give up after 30 seconds.

> > > - at this point the tftp server goes back to the main loop and sees
> > > multiple tftp transfer initiation requests from the aborted transfer
> > > session. 
> > 
> > Initiation requests? You mean WRQ packets from the tftp client you
> > have just killed? 
> 
> Yes.
> 
> I really don't speak PPP, TCP/IP, Windows and tftp well enough to pick
> this apart without more study.
> 
> Perhaps this is something that was lingering in the PPP channel or
> tcp/ip stack... 

That does not make sense. 

We need a trace of packets being transfered to be able to understand
this further. Use windump or something similar on the M$ box to get a
packet capture so we can see what is going on.

        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

[ECOS] Question on wigglers, debugging Redboot in flash for the XScale target in single step mode

[Krishna Ganugapati]

Hello,

I'm looking for some clarification on the following issues.

Target;: Custom IXP425 XScale (almost identical to the IXDP development
board, minus the highspeed UART and minus the LEDs), Redboot in ROM mode.
Debugging interface:  Macraigor Wiggler, OCD Commander for the software
debugger on the host.

Situation:  I can successfully download my redboot rom image to Flash using
the Wiggler and the Macraigor Flash Programmer.
When I reset my target, I connect to the JTAG port using the OCD Commander
debugger and attempt to single step through instructions.

Results

1) Through the OCD Debugger, I note that Flash memory is correctly relocated
at 0x000000000.
2) Using the OCD Debugger, I can manual configure the SDRAM registers and
determine that my SDRAM acc ANy recess is good. I flip the
EXP_BUS_CONFIGURATION bits that switc SDRAM and Flash around and determine
that SDRAM is now located at 0x00000000 - I can successfully read and write
words to the first 256 MB so I am assuming memory is good.

3) Reset the board, board is now in bootup mode, flash is at 0x00000000. I
now attempt to single step through my redboot rom image. On the first
instruction, the board transfers control to the reset handler.

Questions:

 a) Has anyone seen behavior like this?
b) On the web it looks like quite a few people have seen something like
this, but there have been no followups on how to fix this.
c) When I reviewed the XScale documentation, it says that when a debugger
executes a single step instruction, a debug event handler is expected and
that this debug event handler is overloaded at the reset handler. It also
says that the debug event handler code needs to be downloaded to the target
through the JTAG. Questions:
    i) Does the OCD Commander software debugger  interface actually download
a debug event handler? If it does, then my conclusions are wrong and I have
to look elsewhere for why things in single step are not progressing, but if
I'm right, then perhaps I need a more sophisticated debugger for my JTAG
interface (the OCD Commander is free). Any recommendations

Other clarifications:
1) One thing that was not very clear in XScale little endian vs big endian
modes was the opcode fetch modes for the XScale. After investigating, my
understanding is that whether in big endian or in little endian mode,
XScale opcode fetches are always in little endian mode. Which is why when
you build a big endian target for the XScale IXP425 board which generates
the instructions in big endian mode, you need to byte swap the image. Is
this correct?

Thanks for any input.

Krishna


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Question on wigglers, debugging Redboot in flash for the XScale target in single step mode

[Gary Thomas]

On Mon, 2004-08-02 at 08:49, Krishna Ganugapati wrote:
> Hello,
> 
> I'm looking for some clarification on the following issues.
> 
> Target;: Custom IXP425 XScale (almost identical to the IXDP development
> board, minus the highspeed UART and minus the LEDs), Redboot in ROM mode.
> Debugging interface:  Macraigor Wiggler, OCD Commander for the software
> debugger on the host.
> 
> Situation:  I can successfully download my redboot rom image to Flash using
> the Wiggler and the Macraigor Flash Programmer.
> When I reset my target, I connect to the JTAG port using the OCD Commander
> debugger and attempt to single step through instructions.
> 
> Results
> 
> 1) Through the OCD Debugger, I note that Flash memory is correctly relocated
> at 0x000000000.
> 2) Using the OCD Debugger, I can manual configure the SDRAM registers and
> determine that my SDRAM acc ANy recess is good. I flip the
> EXP_BUS_CONFIGURATION bits that switc SDRAM and Flash around and determine
> that SDRAM is now located at 0x00000000 - I can successfully read and write
> words to the first 256 MB so I am assuming memory is good.
> 
> 3) Reset the board, board is now in bootup mode, flash is at 0x00000000. I
> now attempt to single step through my redboot rom image. On the first
> instruction, the board transfers control to the reset handler.
> 
> Questions:
> 
>  a) Has anyone seen behavior like this?
> b) On the web it looks like quite a few people have seen something like
> this, but there have been no followups on how to fix this.
> c) When I reviewed the XScale documentation, it says that when a debugger
> executes a single step instruction, a debug event handler is expected and
> that this debug event handler is overloaded at the reset handler. It also
> says that the debug event handler code needs to be downloaded to the target
> through the JTAG. Questions:
>     i) Does the OCD Commander software debugger  interface actually download
> a debug event handler? If it does, then my conclusions are wrong and I have
> to look elsewhere for why things in single step are not progressing, but if
> I'm right, then perhaps I need a more sophisticated debugger for my JTAG
> interface (the OCD Commander is free). Any recommendations

Why do you think this is wrong?  When you reset the board, single 
stepping (via hardware which is very different that what you describe in
item "c" above), you should execute the "reset handler" an instruction
at a time.  This is just what is supposed to happen.

> 
> Other clarifications:
> 1) One thing that was not very clear in XScale little endian vs big endian
> modes was the opcode fetch modes for the XScale. After investigating, my
> understanding is that whether in big endian or in little endian mode,
> XScale opcode fetches are always in little endian mode. Which is why when
> you build a big endian target for the XScale IXP425 board which generates
> the instructions in big endian mode, you need to byte swap the image. Is
> this correct?
> 
> Thanks for any input.
> 
> Krishna
-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Question on wigglers, debugging Redboot in flash forthe XScale target in single step mode

[Krishna Ganugapati]

Gary, Thanks for the response.

I'm afraid I don't quite understand what you're saying..

a)  > Why do you think this is wrong?  When you reset the board, single
> stepping (via hardware which is very different that what you describe in
> item "c" above), you should execute the "reset handler" an instruction
> at a time.  This is just what is supposed to happen.

Are you saying that this is correct behavior?

When I hit the reset vector, I'm stuck.. subsequent single steps only keep
me at the reset vector - I don't advance any further..

b) How do I debug my redboot image?
-------------------------------------------------
----- Original Message ----- 
From: "Gary Thomas" <gary@mlbassoc.com>
To: "Krishna Ganugapati" <krishnag@marakicorp.com>
Cc: <ecos-discuss@sources.redhat.com>
Sent: Monday, August 02, 2004 7:59 AM
Subject: Re: [ECOS] Question on wigglers, debugging Redboot in flash forthe
XScale target in single step mode


> On Mon, 2004-08-02 at 08:49, Krishna Ganugapati wrote:
> > Hello,
> >
> > I'm looking for some clarification on the following issues.
> >
> > Target;: Custom IXP425 XScale (almost identical to the IXDP development
> > board, minus the highspeed UART and minus the LEDs), Redboot in ROM
mode.
> > Debugging interface:  Macraigor Wiggler, OCD Commander for the software
> > debugger on the host.
> >
> > Situation:  I can successfully download my redboot rom image to Flash
using
> > the Wiggler and the Macraigor Flash Programmer.
> > When I reset my target, I connect to the JTAG port using the OCD
Commander
> > debugger and attempt to single step through instructions.
> >
> > Results
> >
> > 1) Through the OCD Debugger, I note that Flash memory is correctly
relocated
> > at 0x000000000.
> > 2) Using the OCD Debugger, I can manual configure the SDRAM registers
and
> > determine that my SDRAM acc ANy recess is good. I flip the
> > EXP_BUS_CONFIGURATION bits that switc SDRAM and Flash around and
determine
> > that SDRAM is now located at 0x00000000 - I can successfully read and
write
> > words to the first 256 MB so I am assuming memory is good.
> >
> > 3) Reset the board, board is now in bootup mode, flash is at 0x00000000.
I
> > now attempt to single step through my redboot rom image. On the first
> > instruction, the board transfers control to the reset handler.
> >
> > Questions:
> >
> >  a) Has anyone seen behavior like this?
> > b) On the web it looks like quite a few people have seen something like
> > this, but there have been no followups on how to fix this.
> > c) When I reviewed the XScale documentation, it says that when a
debugger
> > executes a single step instruction, a debug event handler is expected
and
> > that this debug event handler is overloaded at the reset handler. It
also
> > says that the debug event handler code needs to be downloaded to the
target
> > through the JTAG. Questions:
> >     i) Does the OCD Commander software debugger  interface actually
download
> > a debug event handler? If it does, then my conclusions are wrong and I
have
> > to look elsewhere for why things in single step are not progressing, but
if
> > I'm right, then perhaps I need a more sophisticated debugger for my JTAG
> > interface (the OCD Commander is free). Any recommendations
>
> Why do you think this is wrong?  When you reset the board, single
> stepping (via hardware which is very different that what you describe in
> item "c" above), you should execute the "reset handler" an instruction
> at a time.  This is just what is supposed to happen.
>
> >
> > Other clarifications:
> > 1) One thing that was not very clear in XScale little endian vs big
endian
> > modes was the opcode fetch modes for the XScale. After investigating, my
> > understanding is that whether in big endian or in little endian mode,
> > XScale opcode fetches are always in little endian mode. Which is why
when
> > you build a big endian target for the XScale IXP425 board which
generates
> > the instructions in big endian mode, you need to byte swap the image. Is
> > this correct?
> >
> > Thanks for any input.
> >
> > Krishna
> -- 
> Gary Thomas <gary@mlbassoc.com>
> MLB Associates
>


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Question on wigglers, debugging Redboot in flash forthe XScale target in single step mode

[Gary Thomas]

On Mon, 2004-08-02 at 09:12, Krishna Ganugapati wrote:
> Gary, Thanks for the response.
> 
> I'm afraid I don't quite understand what you're saying..
> 
> a)  > Why do you think this is wrong?  When you reset the board, single
> > stepping (via hardware which is very different that what you describe in
> > item "c" above), you should execute the "reset handler" an instruction
> > at a time.  This is just what is supposed to happen.
> 
> Are you saying that this is correct behavior?

Yes.

> 
> When I hit the reset vector, I'm stuck.. subsequent single steps only keep
> me at the reset vector - I don't advance any further..
> 
> b) How do I debug my redboot image?

Figure out the answer to why single step doesn't progress, then you'll 
be able to debug RedBoot.  You _should_ be able to do this with the
hardware you have, but alas you may have problems.  Given that MacRaigor
doesn't really support OCD and the Wiggler is pretty primitive and ...

You might consider getting a more powerful debug tool, especially if 
debugging hardware [initialization] is new to you.  One of the best
available is the BDI2000 from Abatron, but there are many others.

> -------------------------------------------------
> ----- Original Message ----- 
> From: "Gary Thomas" <gary@mlbassoc.com>
> To: "Krishna Ganugapati" <krishnag@marakicorp.com>
> Cc: <ecos-discuss@sources.redhat.com>
> Sent: Monday, August 02, 2004 7:59 AM
> Subject: Re: [ECOS] Question on wigglers, debugging Redboot in flash forthe
> XScale target in single step mode
> 
> 
> > On Mon, 2004-08-02 at 08:49, Krishna Ganugapati wrote:
> > > Hello,
> > >
> > > I'm looking for some clarification on the following issues.
> > >
> > > Target;: Custom IXP425 XScale (almost identical to the IXDP development
> > > board, minus the highspeed UART and minus the LEDs), Redboot in ROM
> mode.
> > > Debugging interface:  Macraigor Wiggler, OCD Commander for the software
> > > debugger on the host.
> > >
> > > Situation:  I can successfully download my redboot rom image to Flash
> using
> > > the Wiggler and the Macraigor Flash Programmer.
> > > When I reset my target, I connect to the JTAG port using the OCD
> Commander
> > > debugger and attempt to single step through instructions.
> > >
> > > Results
> > >
> > > 1) Through the OCD Debugger, I note that Flash memory is correctly
> relocated
> > > at 0x000000000.
> > > 2) Using the OCD Debugger, I can manual configure the SDRAM registers
> and
> > > determine that my SDRAM acc ANy recess is good. I flip the
> > > EXP_BUS_CONFIGURATION bits that switc SDRAM and Flash around and
> determine
> > > that SDRAM is now located at 0x00000000 - I can successfully read and
> write
> > > words to the first 256 MB so I am assuming memory is good.
> > >
> > > 3) Reset the board, board is now in bootup mode, flash is at 0x00000000.
> I
> > > now attempt to single step through my redboot rom image. On the first
> > > instruction, the board transfers control to the reset handler.
> > >
> > > Questions:
> > >
> > >  a) Has anyone seen behavior like this?
> > > b) On the web it looks like quite a few people have seen something like
> > > this, but there have been no followups on how to fix this.
> > > c) When I reviewed the XScale documentation, it says that when a
> debugger
> > > executes a single step instruction, a debug event handler is expected
> and
> > > that this debug event handler is overloaded at the reset handler. It
> also
> > > says that the debug event handler code needs to be downloaded to the
> target
> > > through the JTAG. Questions:
> > >     i) Does the OCD Commander software debugger  interface actually
> download
> > > a debug event handler? If it does, then my conclusions are wrong and I
> have
> > > to look elsewhere for why things in single step are not progressing, but
> if
> > > I'm right, then perhaps I need a more sophisticated debugger for my JTAG
> > > interface (the OCD Commander is free). Any recommendations
> >
> > Why do you think this is wrong?  When you reset the board, single
> > stepping (via hardware which is very different that what you describe in
> > item "c" above), you should execute the "reset handler" an instruction
> > at a time.  This is just what is supposed to happen.
> >
> > >
> > > Other clarifications:
> > > 1) One thing that was not very clear in XScale little endian vs big
> endian
> > > modes was the opcode fetch modes for the XScale. After investigating, my
> > > understanding is that whether in big endian or in little endian mode,
> > > XScale opcode fetches are always in little endian mode. Which is why
> when
> > > you build a big endian target for the XScale IXP425 board which
> generates
> > > the instructions in big endian mode, you need to byte swap the image. Is
> > > this correct?
> > >
> > > Thanks for any input.
> > >
> > > Krishna
> > -- 
> > Gary Thomas <gary@mlbassoc.com>
> > MLB Associates
> >
-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Question on wigglers, debugging Redboot in flash fortheXScale target in single step mode

[Krishna Ganugapati]

Thanks Gary. I'm investigating why the single step fails, (I've sent logs to
Macraigor). Thanks for the tip on BDI2000 - I hope to switch debuggers and
JTAG devices.

-- Krishna
-------------------------------------------------
----- Original Message ----- 
From: "Gary Thomas" <gary@mlbassoc.com>
To: "Krishna Ganugapati" <krishnag@marakicorp.com>
Cc: <ecos-discuss@sources.redhat.com>
Sent: Monday, August 02, 2004 8:20 AM
Subject: Re: [ECOS] Question on wigglers, debugging Redboot in flash
fortheXScale target in single step mode


> On Mon, 2004-08-02 at 09:12, Krishna Ganugapati wrote:
> > Gary, Thanks for the response.
> >
> > I'm afraid I don't quite understand what you're saying..
> >
> > a)  > Why do you think this is wrong?  When you reset the board, single
> > > stepping (via hardware which is very different that what you describe
in
> > > item "c" above), you should execute the "reset handler" an instruction
> > > at a time.  This is just what is supposed to happen.
> >
> > Are you saying that this is correct behavior?
>
> Yes.
>
> >
> > When I hit the reset vector, I'm stuck.. subsequent single steps only
keep
> > me at the reset vector - I don't advance any further..
> >
> > b) How do I debug my redboot image?
>
> Figure out the answer to why single step doesn't progress, then you'll
> be able to debug RedBoot.  You _should_ be able to do this with the
> hardware you have, but alas you may have problems.  Given that MacRaigor
> doesn't really support OCD and the Wiggler is pretty primitive and ...
>
> You might consider getting a more powerful debug tool, especially if
> debugging hardware [initialization] is new to you.  One of the best
> available is the BDI2000 from Abatron, but there are many others.
>
> > -------------------------------------------------
> > ----- Original Message ----- 
> > From: "Gary Thomas" <gary@mlbassoc.com>
> > To: "Krishna Ganugapati" <krishnag@marakicorp.com>
> > Cc: <ecos-discuss@sources.redhat.com>
> > Sent: Monday, August 02, 2004 7:59 AM
> > Subject: Re: [ECOS] Question on wigglers, debugging Redboot in flash
forthe
> > XScale target in single step mode
> >
> >
> > > On Mon, 2004-08-02 at 08:49, Krishna Ganugapati wrote:
> > > > Hello,
> > > >
> > > > I'm looking for some clarification on the following issues.
> > > >
> > > > Target;: Custom IXP425 XScale (almost identical to the IXDP
development
> > > > board, minus the highspeed UART and minus the LEDs), Redboot in ROM
> > mode.
> > > > Debugging interface:  Macraigor Wiggler, OCD Commander for the
software
> > > > debugger on the host.
> > > >
> > > > Situation:  I can successfully download my redboot rom image to
Flash
> > using
> > > > the Wiggler and the Macraigor Flash Programmer.
> > > > When I reset my target, I connect to the JTAG port using the OCD
> > Commander
> > > > debugger and attempt to single step through instructions.
> > > >
> > > > Results
> > > >
> > > > 1) Through the OCD Debugger, I note that Flash memory is correctly
> > relocated
> > > > at 0x000000000.
> > > > 2) Using the OCD Debugger, I can manual configure the SDRAM
registers
> > and
> > > > determine that my SDRAM acc ANy recess is good. I flip the
> > > > EXP_BUS_CONFIGURATION bits that switc SDRAM and Flash around and
> > determine
> > > > that SDRAM is now located at 0x00000000 - I can successfully read
and
> > write
> > > > words to the first 256 MB so I am assuming memory is good.
> > > >
> > > > 3) Reset the board, board is now in bootup mode, flash is at
0x00000000.
> > I
> > > > now attempt to single step through my redboot rom image. On the
first
> > > > instruction, the board transfers control to the reset handler.
> > > >
> > > > Questions:
> > > >
> > > >  a) Has anyone seen behavior like this?
> > > > b) On the web it looks like quite a few people have seen something
like
> > > > this, but there have been no followups on how to fix this.
> > > > c) When I reviewed the XScale documentation, it says that when a
> > debugger
> > > > executes a single step instruction, a debug event handler is
expected
> > and
> > > > that this debug event handler is overloaded at the reset handler. It
> > also
> > > > says that the debug event handler code needs to be downloaded to the
> > target
> > > > through the JTAG. Questions:
> > > >     i) Does the OCD Commander software debugger  interface actually
> > download
> > > > a debug event handler? If it does, then my conclusions are wrong and
I
> > have
> > > > to look elsewhere for why things in single step are not progressing,
but
> > if
> > > > I'm right, then perhaps I need a more sophisticated debugger for my
JTAG
> > > > interface (the OCD Commander is free). Any recommendations
> > >
> > > Why do you think this is wrong?  When you reset the board, single
> > > stepping (via hardware which is very different that what you describe
in
> > > item "c" above), you should execute the "reset handler" an instruction
> > > at a time.  This is just what is supposed to happen.
> > >
> > > >
> > > > Other clarifications:
> > > > 1) One thing that was not very clear in XScale little endian vs big
> > endian
> > > > modes was the opcode fetch modes for the XScale. After
investigating, my
> > > > understanding is that whether in big endian or in little endian
mode,
> > > > XScale opcode fetches are always in little endian mode. Which is why
> > when
> > > > you build a big endian target for the XScale IXP425 board which
> > generates
> > > > the instructions in big endian mode, you need to byte swap the
image. Is
> > > > this correct?
> > > >
> > > > Thanks for any input.
> > > >
> > > > Krishna
> > > -- 
> > > Gary Thomas <gary@mlbassoc.com>
> > > MLB Associates
> > >
> -- 
> Gary Thomas <gary@mlbassoc.com>
> MLB Associates
>
>
> -- 
> Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
> and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss
>


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Question on wigglers, debugging Redboot in flash fortheXScale target in single step mode

[Krishna Ganugapati]

Hello Gary and others.

I figured out why my OCD Commander single step was not working... I thoight
I'd post this for anyone else who might be planning to debug rom code
through OCD Commander

From Macraigor:

As I've noted, I can upload my flash memory without trouble, I can dump
disassemble and write data to both flash and memory, which indicates my
internal bus is fine. On single step I get a target bus error 25:11 and I do
not know how to interpret it.

    You need to move the minIcache (containing the JTAG monitor) out of the
way of your flash . By
     default the miniIcache starts out mapped to 0x0 - 2K, You need to enter
on the OCD Commander's
    command line :

        "xmon 0x500000"   (valid address values are 0 - 0x1000000)
    this will move the miniIcache to an area that is not going to be
executed by your load image.  You  should then
     be able to step and run without a problem.

Once I did this, it worked beautifully. Sadly, after several dozen
instructions, in the process of attempting the first store operation, it
fails with a Read Large: Target Bus Error 41:11. The interesting thing here
is this is the first set of instructions that attempt to write the
Configuration Register values. Prior to this all the instructions are about
setting the endianness (from default little to big), invalidating the
caches, configuring the MMU etc. All these prior instructions  manipulate
data between the registers, the coprocessor regisrters and some load
immediate instructions. The instructions that forces the error is a str r0,
[r1] where both r0 and r1 are good.

I asked Macraigor what this is - still waiting to hear from them....


-------------------------------------------------
----- Original Message ----- 
From: "Gary Thomas" <gary@mlbassoc.com>
To: "Krishna Ganugapati" <krishnag@marakicorp.com>
Cc: <ecos-discuss@sources.redhat.com>
Sent: Monday, August 02, 2004 8:20 AM
Subject: Re: [ECOS] Question on wigglers, debugging Redboot in flash
fortheXScale target in single step mode


> On Mon, 2004-08-02 at 09:12, Krishna Ganugapati wrote:
> > Gary, Thanks for the response.
> >
> > I'm afraid I don't quite understand what you're saying..
> >
> > a)  > Why do you think this is wrong?  When you reset the board, single
> > > stepping (via hardware which is very different that what you describe
in
> > > item "c" above), you should execute the "reset handler" an instruction
> > > at a time.  This is just what is supposed to happen.
> >
> > Are you saying that this is correct behavior?
>
> Yes.
>
> >
> > When I hit the reset vector, I'm stuck.. subsequent single steps only
keep
> > me at the reset vector - I don't advance any further..
> >
> > b) How do I debug my redboot image?
>
> Figure out the answer to why single step doesn't progress, then you'll
> be able to debug RedBoot.  You _should_ be able to do this with the
> hardware you have, but alas you may have problems.  Given that MacRaigor
> doesn't really support OCD and the Wiggler is pretty primitive and ...
>
> You might consider getting a more powerful debug tool, especially if
> debugging hardware [initialization] is new to you.  One of the best
> available is the BDI2000 from Abatron, but there are many others.
>
> > -------------------------------------------------
> > ----- Original Message ----- 
> > From: "Gary Thomas" <gary@mlbassoc.com>
> > To: "Krishna Ganugapati" <krishnag@marakicorp.com>
> > Cc: <ecos-discuss@sources.redhat.com>
> > Sent: Monday, August 02, 2004 7:59 AM
> > Subject: Re: [ECOS] Question on wigglers, debugging Redboot in flash
forthe
> > XScale target in single step mode
> >
> >
> > > On Mon, 2004-08-02 at 08:49, Krishna Ganugapati wrote:
> > > > Hello,
> > > >
> > > > I'm looking for some clarification on the following issues.
> > > >
> > > > Target;: Custom IXP425 XScale (almost identical to the IXDP
development
> > > > board, minus the highspeed UART and minus the LEDs), Redboot in ROM
> > mode.
> > > > Debugging interface:  Macraigor Wiggler, OCD Commander for the
software
> > > > debugger on the host.
> > > >
> > > > Situation:  I can successfully download my redboot rom image to
Flash
> > using
> > > > the Wiggler and the Macraigor Flash Programmer.
> > > > When I reset my target, I connect to the JTAG port using the OCD
> > Commander
> > > > debugger and attempt to single step through instructions.
> > > >
> > > > Results
> > > >
> > > > 1) Through the OCD Debugger, I note that Flash memory is correctly
> > relocated
> > > > at 0x000000000.
> > > > 2) Using the OCD Debugger, I can manual configure the SDRAM
registers
> > and
> > > > determine that my SDRAM acc ANy recess is good. I flip the
> > > > EXP_BUS_CONFIGURATION bits that switc SDRAM and Flash around and
> > determine
> > > > that SDRAM is now located at 0x00000000 - I can successfully read
and
> > write
> > > > words to the first 256 MB so I am assuming memory is good.
> > > >
> > > > 3) Reset the board, board is now in bootup mode, flash is at
0x00000000.
> > I
> > > > now attempt to single step through my redboot rom image. On the
first
> > > > instruction, the board transfers control to the reset handler.
> > > >
> > > > Questions:
> > > >
> > > >  a) Has anyone seen behavior like this?
> > > > b) On the web it looks like quite a few people have seen something
like
> > > > this, but there have been no followups on how to fix this.
> > > > c) When I reviewed the XScale documentation, it says that when a
> > debugger
> > > > executes a single step instruction, a debug event handler is
expected
> > and
> > > > that this debug event handler is overloaded at the reset handler. It
> > also
> > > > says that the debug event handler code needs to be downloaded to the
> > target
> > > > through the JTAG. Questions:
> > > >     i) Does the OCD Commander software debugger  interface actually
> > download
> > > > a debug event handler? If it does, then my conclusions are wrong and
I
> > have
> > > > to look elsewhere for why things in single step are not progressing,
but
> > if
> > > > I'm right, then perhaps I need a more sophisticated debugger for my
JTAG
> > > > interface (the OCD Commander is free). Any recommendations
> > >
> > > Why do you think this is wrong?  When you reset the board, single
> > > stepping (via hardware which is very different that what you describe
in
> > > item "c" above), you should execute the "reset handler" an instruction
> > > at a time.  This is just what is supposed to happen.
> > >
> > > >
> > > > Other clarifications:
> > > > 1) One thing that was not very clear in XScale little endian vs big
> > endian
> > > > modes was the opcode fetch modes for the XScale. After
investigating, my
> > > > understanding is that whether in big endian or in little endian
mode,
> > > > XScale opcode fetches are always in little endian mode. Which is why
> > when
> > > > you build a big endian target for the XScale IXP425 board which
> > generates
> > > > the instructions in big endian mode, you need to byte swap the
image. Is
> > > > this correct?
> > > >
> > > > Thanks for any input.
> > > >
> > > > Krishna
> > > -- 
> > > Gary Thomas <gary@mlbassoc.com>
> > > MLB Associates
> > >
> -- 
> Gary Thomas <gary@mlbassoc.com>
> MLB Associates
>
>
> -- 
> Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
> and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss
>


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Øyvind Harboe]

I haven't got the setup available right now, but I think I remember the steps 
to reproduce a bit better:

1. start tftp transfer
2. abort tftp transfer
3. at this point the new tftp server thread is waiting for timeout on
#1
4. start new tftp transfer
5. wait a bit(but not long enough for tftp transfer to start).
6. abort tftp transfer
7. at this point the tftp server is stuck in the main loop timing out on
WRQ requests. Each WRQ request requires a tftpd_write_file() to time out, so
this takes a loooong time.

My working theory is that the Windows tftp client, in #4,
filled up the socket with WRQ requests while it was trying 
to get in touch with the tftp server. Closing and opening the sockets,
which happens as an unintended(?) side effect of setting
CYGSEM_NET_TFTPD_MULTITHREADED=1, flushes any such obsolete
WRQ requests.

Once I some time and my hands on the system again, I'll run the 
above tests and try to capture some packages. I'll also try to remove
Windows from the equation by reproducing the problem with a Linux
tftp client.




-- 
Øyvind Harboe
http://www.zylin.com



--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Andrew Lunn]

On Tue, Aug 03, 2004 at 08:40:46AM +0200, ?yvind Harboe wrote:
> I haven't got the setup available right now, but I think I remember the steps 
> to reproduce a bit better:
> 
> 1. start tftp transfer
> 2. abort tftp transfer
> 3. at this point the new tftp server thread is waiting for timeout on
> #1
> 4. start new tftp transfer
> 5. wait a bit(but not long enough for tftp transfer to start).
> 6. abort tftp transfer
> 7. at this point the tftp server is stuck in the main loop timing out on
> WRQ requests. Each WRQ request requires a tftpd_write_file() to time out, so
> this takes a loooong time.

This now make sense. It is the starting of the new connection before
the old one times out that causes the delay. Since the system is
single threaded it has two choices when the new connection request
arrives. The current implementation simply buffers the request until
later. The alternative implementation would be to discard the request
and let the client retry. Both implementations are probably correct
but have pros and cons. All these aborted connections is a corner case
that should not happen very often so i would not worry about it.

Having said that, i think the real solution is to teach your users to
have some patients and not abort the connections....

        Andrew

-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

Re: [ECOS] Problems when aborting tftp transfers

[Øyvind Harboe]

On Tue, 2004-08-03 at 11:49, Andrew Lunn wrote:
> On Tue, Aug 03, 2004 at 08:40:46AM +0200, ?yvind Harboe wrote:
> > I haven't got the setup available right now, but I think I remember the steps 
> > to reproduce a bit better:
> > 
> > 1. start tftp transfer
> > 2. abort tftp transfer
> > 3. at this point the new tftp server thread is waiting for timeout on
> > #1
> > 4. start new tftp transfer
> > 5. wait a bit(but not long enough for tftp transfer to start).
> > 6. abort tftp transfer
> > 7. at this point the tftp server is stuck in the main loop timing out on
> > WRQ requests. Each WRQ request requires a tftpd_write_file() to time out, so
> > this takes a loooong time.
> 
> This now make sense. It is the starting of the new connection before
> the old one times out that causes the delay. Since the system is
> single threaded it has two choices when the new connection request
> arrives. The current implementation simply buffers the request until
> later. The alternative implementation would be to discard the request
> and let the client retry. Both implementations are probably correct
> but have pros and cons. All these aborted connections is a corner case
> that should not happen very often so i would not worry about it.
> 
> Having said that, i think the real solution is to teach your users to
> have some patients and not abort the connections....

:-)

The problem is that the timeout is *very* long because it is number of 
queued WRQ (ca 5) * timeout for tftpd_write_file().

A single timeout is acceptable, but minutes is very confusing for the
user.

>         Andrew

-- 
Øyvind Harboe
http://www.zylin.com



--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss