> Here is the list of known problems I am aware of and of fixes not yet committed
> to midas svn:
> 
> 1) added variable /equiment/foo/common/PerVariableHistory

corrected in svn revision 4203, read
http://savannah.psi.ch/viewcvs/trunk/src/mlogger.c?root=midas&rev=4203&sortby=rev&view=log

> 2) writing compressed midas files (foo.mid.gz) crashes the mlogger when file
> size reaches 2 GBytes. This problem could be new in SL5.1.

(no change)

> 3) when a midas client becomes unresponsive, runs cannot be stopped using the
> "stop" button in mhttpd. This is because cm_transition() loops over all attached
> clients, but never removes clients that are known to be dead. Proposed fix is to
> call cm_check_client() for each client before calling their rpc transition handler.

Fixed in SVN revision 4198, read
http://savannah.psi.ch/viewcvs/trunk/src/midas.c?root=midas&rev=4201&sortby=rev&view=log

> 4) the discussed before fix for reading broken history files (skip bad data).

Fixed in SVN revision 4202, read https://ladd00.triumf.ca/elog/Midas/482

> 5) mhttpd history "export" button needs to be fixed (by request from ALPHA). At
> present it either does not return all exiting data or crashes mhttpd. (no fix)

(no change)

> 6) mhttpd ODB editor in "set value" page, the "cancel" button is broken (needs
> to be corrected for "relative URL").

Apply this patch to src/mhttpd.c

@@ -11156,10 +11190,7 @@
          sprintf(str, "SC/%s/%s", eq_name, group);
          redirect(str);
       } else {
-         strlcpy(str, path, sizeof(str));
-         if (strrchr(str, '/'))
-            strlcpy(str, strrchr(str, '/')+1, sizeof(str));
-         redirect(str);
+         redirect("./");
       }

> 7) mhttpd needs AJAX-style methods for reading and writing ODB. (no fix)

(no change)

K.O.

> > Here is the list of known problems I am aware of and of fixes not yet committed to midas svn:
> > 1) added variable /equiment/foo/common/PerVariableHistory
> 
> corrected in svn revision 4203, read
> http://savannah.psi.ch/viewcvs/trunk/src/mlogger.c?root=midas&rev=4203&sortby=rev&view=log

Was still broken - all should work in revision 4207.

> > 2) writing compressed midas files (foo.mid.gz) crashes the mlogger when file
> > size reaches 2 GBytes. This problem could be new in SL5.1.

It turns out that on SL5 and SL5.1 (and others?) the 32-bit version of ZLIB opens the
compressed output file without the O_LARGEFILE flag, this limits the file size to 2 GB.

Fixed by opening the file ourselves, then attach compression stream using gzdopen().

Revision 4207. (Not tested on Windows - may be broken!)

> > 5) mhttpd history "export" button needs to be fixed (by request from ALPHA). At
> > present it either does not return all exiting data or crashes mhttpd. (no fix)
> 
> (no change)
> 
> > 6) mhttpd ODB editor in "set value" page, the "cancel" button is broken (needs
> > to be corrected for "relative URL").
> 
> Apply this patch to src/mhttpd.c
> 
> @@ -11156,10 +11190,7 @@
>           sprintf(str, "SC/%s/%s", eq_name, group);
>           redirect(str);
>        } else {
> -         strlcpy(str, path, sizeof(str));
> -         if (strrchr(str, '/'))
> -            strlcpy(str, strrchr(str, '/')+1, sizeof(str));
> -         redirect(str);
> +         redirect("./");
>        }
> 
> > 7) mhttpd needs AJAX-style methods for reading and writing ODB. (no fix)
> 
> (no change)
> 
> K.O.

> The current versions of mlogger SVN 4215 is flooding our message system with
> stuff like
> 
> > Tue Jun 10 16:42:01 2008 [Logger,INFO] Configured history with 22 events
> > Tue Jun 10 16:42:14 2008 [Logger,INFO] Configured history with 22 events
> > Tue Jun 10 16:42:26 2008 [Logger,INFO] Configured history with 22 events
> 
> This is fatal to us and blowing up the midas.log like hell. I would prefer if
> one could flag these kind of messages (ODB /Logger/..), i.e. enable and disable
> it. At the moment I have to comment it out in the source code since we cannot
> work with it.

I just sent the attached message to Stefan - please read it.

Before we take any action, we need to understand why history is being
reconfigured every 10 seconds (according to your logfile snippet).

Are you starting a new run every 10 seconds?

If that is what you do and that is your intent, I guess it is atypical usage of
MIDAS and the message from the mlogger is offensive and should be removed/disabled.

If something else is going on, we need to understand it before we sweep trouble
under the carpet by disabling this message.

K.O.

Stefan - there is more bad news - the message is produced when the history
is being reconfigured. This only is supposed to happen when the mlogger
starts or at the begin of run.

So these messages are just a tip of an iceberg of some other trouble.

The logic of when history is reconfigured I did not change. So likely
the trouble existed before, but you did not know about it.

We can kill the message, but why is the history being reconfigured
at a rate that "floods the log file"? That cannot possibly be good.

K.O.

> The current versions of mlogger SVN 4215 is flooding our message system with
> stuff like
> 
> > Tue Jun 10 16:42:01 2008 [Logger,INFO] Configured history with 22 events
> > Tue Jun 10 16:42:14 2008 [Logger,INFO] Configured history with 22 events
> > Tue Jun 10 16:42:26 2008 [Logger,INFO] Configured history with 22 events

Problem confirmed on the M11 DAQ system at TRIUMF. We definitely do nothing funny
there, so what is going on? Will investigate.

K.O.

Following changes to midas produced from the TRIUMF T2K project have been
committed to svn:
1) cm_shutdown() will now SIGKILL clients that cannot be stopped via normal
means. Previously cm_shutdown() would print a message to the effect "please kill
this client yourself manually". The user action in this case (assuming they did
not issue cm_shutdown() by mistake) has been to find out the client pid using
"ps", kill -KILL it, then "odbedit clean". cm_shutdown() now performs all this
automatically.
2) rpc_send_event() did not correctly detect loss of connection to the remote
mserver (i.e. in case it was killed by cm_shutdown() above). Now, correct error
handling is in place and the remote frontend should gracefully shutdown if
mserver connection is lost. (However I observe that some of my remote frontends
fail to exit unless I do "exit(1);" from my frontend_exit() function.
3) mhttpd bug fixed: when editing odb entries, the "cancel" button did not work
correctly.
4) lazylogger "script" backup type is now fully tested and documented. Example
scripts for writing to dcache are available by request.
5) mlogger and mhttpd changes for writing history data to an sql database are
mostly completed and will be committed after some more debugging. (If you are
interested in details, please contact me directly).
6) (committed some time ago) Makefile changes for cross-compiling midas are now
in: "make linux32", "make linux64", "make crosscompile".
K.O.

The code for storing midas history in an odbc sql database has been committed.
Changes:
include/history_odbc.h, src/history_odbc.cxx --- implementation
src/mlogger.c --- call the history_odbc functions
utils/mh2sql.cxx --- import existing midas history files (*.hst) into an odbc
sql database.

This new code is enabled by the HAVE_ODBC gunk in the Makefile. If compilation
bombs, please let me know and as a work around, comment out all instances of
HAVE_ODBC from your Makefile.

Limitations:
- mhttpd support for reading history data from odbc sql database is missing
- many sql functions are implemented in a very minimalistic form (i.e. when
defining a history event, we blindly ask sql to create the tables, even if they
already exist - this works, but spams the midas log with sql errors).
- error handling is incomplete: after any sql error, the odbc connection is closed.
- only MySQL (and ascii output) are supported: we use mysql-specific data types
as they match midas types exactly. Code to support PgSQL is present and it used
to work, but is commented out. (At TRIUMF/T2K, we intend to use MySQL exclusively).
- ODBC ascii interface is used, instead of the potentially more efficient binary
interface.

To enable:
- create a MySQL database,
- create $HOME/.odbc.ini (see attached example)
- set ODB "/History/PerVariableHistory" to "1" - the new code is intended to be
used with per-variable history. Per-equipment (traditional) history would work,
but will result in suboptimal layout of SQL tables.
- set ODB "/Logger/ODBC_DSN" to the DSN defined in .odc.ini.
- set ODB "/Logger/ODBC_Debug" to non-zero to enable debugging output from the
new code.

To use the "ascii output" mode:
Included is code to write "ascii" sql output into a text file, instead of using
an actual SQL database. To enable it, set "ODBC_DSN" to
"/path/to/some/text/file" and all SQL output will be written to this file. No
actual SQL database required. This mode exists mostly for debugging the SQL syntax.

Despite limitations, the committed code is fully functional - we are presently
using it to record history data from slow controls of T2K detector tests
(voltages, currents, temperatures).

Comments and suggestions on naming and mapping from odb structures to SQL tables
is very much welcome.

K.O.

midas revision 4342 implements non-default tcp port numbers for the mserver.

To use, run "mserver -p 7070" and say "setenv MIDAS_SERVER_HOST
host.example.com:7070".

This is useful when multiple experiments share the same computer, but one does
not want to setup a global /etc/exptab (non-root users cannot change it) or one
does not want to run the mserver from xinetd (i.e. all experiments run different
versions of midas and cannot use the same common mserver executable).

Changed files:
src/mserver.c
src/midas.c
doxfiles/utilities.dox
doxfiles/appendixD.dox

Revision 4342.

K.O.

mhttpd "messages" page stopped working after svn revision 4327 because of uninitialized variable 
"filename2" in midas.c:cm_message_retrieve(). Attached patch fixes the problem for me.
K.O.


--- src/midas.c (revision 4342)
+++ src/midas.c (working copy)
@@ -978,6 +978,8 @@
       size = sizeof(filename);
       db_get_value(hDB, 0, "/Logger/Message file", filename, &size, TID_STRING, TRUE);
 
+      strlcpy(filename2, filename, sizeof(filename2));
+
       if (strchr(filename, '%')) {
          /* replace strings such as midas_%y%m%d.mid with current date */
          tzset();

The latest midas mvmestd.h driver for the Tundra tsi148 pci-vme bridge as used
on GEFANUC VME processors have been commited, revision 4349.

This midas drivers require the "gefvme" Linux kernel driver supplied by GEFANUC
as part of their Linux BSP. (Note that version "v7865-sdk-linux-R01.00" from
GEFANUC is mostly non-functional).

At TRIUMF have the V7865 VME processors and use the kernel driver
v7865-sdk-linux-R01.00-KO6. This driver supports these functions:

1) memory mapped access to full VME A16 and A24 address spaces and window-mapped
access to VME A32 address space. (original gefvme driver does not do
memory-mapped access)
2) DMA directly from vme to user memory, with support for multi-segment chained
transfers (original gefvme driver lacks chained transfers)
3) DMA from user memort to vme should work but is untested
4) no support for interrupts (original gefvme driver does not interrupts).

If you are interested in in using the TRIUMF driver, please contact me directly.

If you already purchased the GEFANUC BSP, I think you can use my drivers
immediately, without objection from GEFANUC.

Otherwise, I will have to do some research into the gefvme code license: since
all of the code appears to have GPL headers and identical code exists on the
internet, I expect to find that my gefvme driver can be freely distributed under
the GPL. But until then, and until it is cleared with TRIUMF management, I
cannot make my gefvme driver available for free download.

K.O.

As we were looking into problems with starting and stopping runs in one of our
daq systems, we found that the mlogger does something differently compared to
mhttpd and odbedit. Starting and stopping runs from mhttpd and odbedit works
correctly, but runs restarted by the file size limit in mlogger would often have
problems.

It turns out that mlogger calls cm_transition() with the ASYNC flag, while
mhttpd and odbedit always use SYNC.

The best I can tell, the ASYNC flag tells cm_transition() to fire off the
end-run rpc calls to all clients all at once, without waiting for reply from the
previous client before calling the next one. This effectively defeats the
transition sequence numbers - higher-numbered clients are told to end-run before
the lower-numbered clients have finished their end-run processing.

Most of the time, transition sequence numbers do not matter - all frontends can
stop at the same time, only mlogger has to be the very last, and for transitions
initiated by the mlogger itself, this sequencing is preserved.

It turns out that for our system, correct sequencing of individual frontends is
important, for example, the frontend controlling the trigger system has to stop
first. As we are using correctly adjusted transition sequence numbers, the right
sequence is always done when runs are started/stopped from mhttpd and from
odbedit, but not for runs started/stopped by the mlogger.

So by changing mlogger to always do SYNC transitions, we fixed our sequencing
problem - now runs always start and stop correctly.

But then we ran into a deadlock between the mlogger and the event builder:

1) mlogger wants to stop the run
2a) mlogger stops reading the SYSTEM buffer
2b) mlogger starts cm_transition(SYNC)
3) rpc call to trigger frontend, trigger is blocked (no new events are
generated, but existing data is still flowing through the system)
4) other frontends are stopped (data still flowing)
5) data still flowing through the system, into the event builder, into the
SYSTEM buffer
6) SYSTEM buffer becomes 100% full (mlogger is not reading it, it is busy inside
cm_transition()), event builder is waiting for free space inside bm_send_event()
7) mlogger issues end-run rpc call to event builder
8) deadlock: mlogger is waiting for a reply from the event builder, the event
builder is waiting for free space in the SYSTEM buffer (not processing rpc
calls), mlogger is supposed to empty the SYSTEM buffer, but it is waiting for an
rpc reply instead.

In our particular case, the dead lock was easy to avoid by making the SYSTEM
buffer big enough to accommodate all in-flight data, but the problem remains in
the general case. I suspect mlogger uses ASYNC transactions exactly to avoid
this type of deadlock (mlogger used ASYNC transactions since svn revision 2, the
beginning of time).

Personally, I am not happy about the inconsistency of run sequencing between
mlogger and mhttpd/odbedit (hmm... should also check mfe.c, it also stops runs
based on event count limits, etc). I think it would be better if all programs
did the same exact thing when starting/stopping runs. When mlogger does
something different, we get surprising unexpected behaviour, best avoided.

One possible solution could be to add an odb variable "/logger/async
transitions", set to "false" by default - to be consistent with other programs.
Systems that benefit from the old ASYNC behaviour and do not care about exact
sequencing can set this flag to "true".

K.O.

The Makefile targets for crosscompiling MIDAS are now documented in the MIDAS
Doxygen documentation:

make linux32 & make clean32
make linux64 & make clean64
make crosscompile
make dox

This has to do with which flavour of MIDAS is built by default: 32-bit or 64-bit.

This is how this works now.

Default flavour is determined by ROOT. If ROOTSYS points to 32-bit ROOT, then
32-bit MIDAS is built, if 64-bit ROOT, then 64-bit MIDAS. This works well after
the ROOT team added the correct "-m32" and "-m64" flags to "rootconfig --cflags".

If for some reason, we also need a non-default flavour of MIDAS, for example
when the main daq computer runs 64-bit MIDAS, but one frontend has to run on a
"32-bit only" VME processor, you say "make linux32". This creates the
"linux-m32/{lib,bin}" tree that you then reference in the Makefile of your
special frontend (i.e. instead of "-L$MIDASSYS/linux/lib" say
"-L$MIDASSYS/linux-m32/lib"). "make linux64" works the same way.

These non-default flavours of MIDAS are compiled with most special features
disabled: no ROOT, no MYSQL, etc.

When building "make linux32", you may also see errors caused by missing 32-bit
libraries - many 64-bit Linux distributions do not install the full 32-bit
development environment by default - so some header files and libraries may be
reported as missing. These not-installed-by-default 32-bit packages are usually
easy to install using commands like "yum install libxxx-devel.i386".

K.O.

A new set of functions was added to mscb.h to adjust mscb timeouts and retries to better match specific 
applications:

+   int EXPRT mscb_get_max_retry();
+   int EXPRT mscb_set_max_retry(int max_retry);
+   int EXPRT mscb_get_usb_timeout();
+   int EXPRT mscb_set_usb_timeout(int timeout);
+   int EXPRT mscb_get_eth_max_retry();
+   int EXPRT mscb_set_eth_max_retry(int eth_max_retry);

There are 3 settings:

1) mscb_max_retry: most (all?) mscb operations, like mscb_read(), retry failed mscb transactions up to 
10 times. The corresponding set and get functions allow tuning this retry limit.

2) mscb_usb_timeout: the driver for the USB-MSCB adapter uses a timeout of 6 seconds. 
mscb_set_usb_timeout() permits changing this value.

3) mscb_eth_max_retry: the driver for the Ethernet-MSCB adapter has to deal with UDP packet loss. If 
the adapter does not respond to a UDP command, the UDP command is sent again, with a bigger 
timeout (timeout = 100 * (retry+1), in ms), this is repeated up to 10 times. mscb_set_eth_max_retry() 
permits adjusting this number of retries.

This is how it works for the usb interface:

int mscb_read(...)
   for (retry=0; retry<mscb_max_retry; retry++)
       mscb_exch()
            musb_write(..., mscb_usb_timeout)
            musb_read(..., mscb_usb_timeout)     

This is how it works for the ethernet interface:

int mscb_read(...)
   for (retry=0; retry<mscb_max_retry; retry++)
       mscb_exch()
            for (retry=0; retry<mscb_eth_max_retry; retry++)
                 send_udp_command()
                 wait_for_udp_response(timeout = 100 * (retry+1))

This is how the new functions are intended to be used:
   ...
   int old = mscb_set_max_retry(2);
   ... do stuff ...
   mscb_set_max_retry(old); // restore default value

svn revision 4356.
K.O.

In midas.c there are several places where client last activity time stamps are checked against the 
watchdog timeout and the clients are declared dead if they fail to update their activity time stamps. 
ODB time stamps and data buffer time stamps appear to be handled in a similar manner.

Most checks are done like this:

now = ss_millitime();
if (client->watchdog > 0      <----- check that the watchdog is enabled
    && now > client->last_activity    <---- check for crazy time stamps from the future
    && now - client->last_activity > client->watchdog_timeout)   <--- normal timeout
        remove_client(client);

But in a few places, the extra checks are missing:

now = ss_millitime();
if (now - client->last_activity > client->watchdog_timeout)
        remove_client(client);

Is this an oversight from when additional checks were added?
Should I make all checks read like the first one?

K.O.

When I run odbedit remotely (odbedit -h ladd09), the "cleanup" command unexpectedly produces the 
output of the "sor" command (sure enough, there is a call to db_get_open_records() there), but when I run 
it locally, I do not get this output (but db_get_open_records() is still called). Strange. K.O.

When mserver receives events from remote client, writes them into a data buffer and this data buffer 
becomes 100% full, we see mserver go into 100% consumption.

It turns out this happens because bm_wait_for_free_space() never sleeps, instead, it busy-loops waiting 
for free space. bm_wait_for_free_space() does call ss_suspend(), but ss_suspend() does not sleep 
because there is pending data in the event network connection and it want to process it.

Best solution I have is to use silly "if (ss_suspend()!=SS_TIMEOUT) sleep(1);"

Also read this explanation: (bm_cleanup is needed to detect that the client holding the buffer at 100% 
full (a stuck or dead GET_ALL reader, mevb in our case), has been killed off and we can continue as 
usual)

       /* signal other clients wait mode */
       pheader->client[bm_validate_client_index(pbuf)].write_wait = requested_space;
 
+      bm_cleanup("bm_wait_for_free_space", ss_millitime(), FALSE);
+
       status = ss_suspend(1000, MSG_BM);
 
+      /* make sure we do sleep in this loop:
+       * if we are the mserver receiving data on the event
+       * socket and the data buffer is full, ss_suspend() will
+       * never sleep: it will detect data on the event channel,
+       * call rpc_server_receive() (recursively, we already *are* in
+       * rpc_server_receive()) and return without sleeping. Result
+       * is a busy loop waiting for free space in data buffer */
+      if (status != SS_TIMEOUT)
+         sleep(1);
+
       /* validate client index: we could have been removed from the buffer */
       pheader->client[bm_validate_client_index(pbuf)].write_wait = 0;

K.O.

Around Wednesday Noon, there was a power outage at triumf (loss of ups power in the triumf 
computing center) and after rebooting ladd00, https/ssl access stopped working with a complaint 
about mismatching server name and ssl certificate name. This configuration used to work, so one of the 
system updated must have broke it. This problem is now fixed and access to midas elog is restored. 
K.O.

It turns out that on 32-bit Linux, ss_file_size() returns the wrong answer for
files bigger than 2 GB (4GB?). The Linux stat() system call returns an error
(which is ignored) and bogus file size data (returned to the caller).

On 64-bit Linux (compiled with -m64), stat() appears to return correct data.

Related functions ss_disk_size() and ss_disk_free() return correct answers on
both 32-bit and 64-bit Linux (biggest disk I tried was 5.5 TB).

I now fixed this problem by using the stat64() system call for "#ifdef OS_LINUX".

I also changed ss_file_size(), ss_disk_size() and ss_disk_free() to return -1 if
the system call returns an error. I also added a test program
utils/test_ss_file_size.c.

svn revision 4397.
K.O.

I now have a better understanding of this: lazylogger uses ss_file_size() to find
out if a file exists or not. This function used to return 0 (probably) for
non-existant files (there was no check for error status from stat() system call,
so the return value for non-existant files was never well defined).

With ss_file_size() returning 0 for nonexistant files, 0-size files clearly cause
problems to lazylogger.

Now, since svn revision 4397, ss_file_size() returns -1 for non-existant files,
but lazylogger still needs to be tought about this.

The problem "lazylogger does not like 0-size files" remains for now.

K.O.


> With latest midas, I see this:
> 
> Thu Oct 14 19:31:17 2004 [Lazy_Tape] [lazylogger.c:1717:Lazy] lazy_file_exists
> file run17567.ybs doesn't exists
> Thu Oct 14 19:31:27 2004 [Lazy_Tape] [lazylogger.c:1717:Lazy] lazy_file_exists
> file run17567.ybs doesn't exists
> 
> The file run17567.ybs has size zero:
> 
> -rw-r--r--    1 twistonl users      950272 Oct 13 19:29
> /twist/data_onl/current/run17565.ybs
> -rw-r--r--    1 twistonl users      950272 Oct 13 19:45
> /twist/data_onl/current/run17566.ybs
> -rw-r--r--    1 twistonl users           0 Oct 13 20:00
> /twist/data_onl/current/run17567.ybs
> -rw-r--r--    1 twistonl users      983040 Oct 13 20:03
> /twist/data_onl/current/run17568.ybs
> -rw-r--r--    1 twistonl users      950272 Oct 13 20:26
> /twist/data_onl/current/run17569.ybs
> 
> I am not sure how to fix this lazylogger logic. Please help.
> 
> K.O.

lazylogger was updated to improve handling of the list of runs still on disk
(odb /Lazy/xxx/List).

Previously, each and every run was listed in the List arrays. With modern
Terabyte-sized data disks, many many days worth of runs tend to remain on disk
and these List arrays were getting too big, inflating the size of ODB dumps
written by mlogger into the output data file and slowing down starting and
stopping of runs considerably.

Now, the runs are listed as ranges of "first run" - "last run", (see example below).

This significantly reduces the size of the "List" arrays and makes lazylogger
usable for the ALPHA experiment at CERN and for T2K/ND280 prototype DAQ at
TRIUMF (writing to Castor and Dcache respectively, using the newly added
"Script" method).

The new List format is fully compatible with the old format and you can update
and run the new lazylogger without changing anything in ODB. New runs will be
added to the List arrays in the new format and data in the old format will
eventually go away as old runs are removed from disk.

svn revision 4394.
K.O.

Example: this reads like this:
range from 7100 to 7154
range from 7157 to 7161 (7155-7156 are missing)
range from 7163 to 7168 (7162 is missing)
runs 7170, 7173, 7176
range from 7179 to 7182
and so forth.

ODB /Lazy/Dcache/List
007100
[0] 7100 (0x1BBC)
[1] -7154 (0xFFFFE40E)
[2] 7157 (0x1BF5)
[3] -7161 (0xFFFFE407)
[4] 7163 (0x1BFB)
[5] -7168 (0xFFFFE400)
[6] 7170 (0x1C02)
[7] 7173 (0x1C05)
[8] 7176 (0x1C08)
[9] 7179 (0x1C0B)
[10] -7182 (0xFFFFE3F2)
[11] 7184 (0x1C10)
[12] 7188 (0x1C14)
[13] -7199 (0xFFFFE3E1)
007200
[0] 7200 (0x1C20)
[1] -7225 (0xFFFFE3C7)

> revision 4142+4143 are minor fixes, refactoring (switch the code to use helper
> functions) and implementation of history for structured banks

The implementation of "history for structured banks" had a bug - tags inside
structured banks were counted incorrectly, leading to memory overwrites and mlogger
crash in open_history().

This is problem is now fixed (plus added assert() checks to crash-out if overwrite of
tags[] array is detected).

svn revision 4398.
K.O.