Difference between revisions of "MAS Bug List"
From MCEWiki
(→device driver) |
|||
| (4 intermediate revisions by 3 users not shown) | |||
| Line 2: | Line 2: | ||
* dsp_ops: write_callback awakens the read sleeper before updating the command state. That leads to occasional dsp_cmd hangups. | * dsp_ops: write_callback awakens the read sleeper before updating the command state. That leads to occasional dsp_cmd hangups. | ||
| + | * dsp_pci/dsp_driver: there are race conditions in the cleanup/remove procedure. Interrupts aren't disabled until after higher levels de-init, meaning callbacks get sent into nowhere. | ||
* Driver doesn't work well on realtime (in NFY/HST mode). Oddly, non-realtime acquisitions at 400 Hz in this mode have been achieved. | * Driver doesn't work well on realtime (in NFY/HST mode). Oddly, non-realtime acquisitions at 400 Hz in this mode have been achieved. | ||
* Inform interval should be remembered, and updated whenever necessary. [ Careful with this, you have to watch for things like DSP resets / mode switches! ] | * Inform interval should be remembered, and updated whenever necessary. [ Careful with this, you have to watch for things like DSP resets / mode switches! ] | ||
| − | |||
* Incorrect DSP packet size leads to DSP dropping the packet and no one knows to time out. | * Incorrect DSP packet size leads to DSP dropping the packet and no one knows to time out. | ||
* bigphysarea causes system hang when driver calls it and "bigphysarea=NNNN" hasn't been given as a kernel option. This is bigphysarea's fault, probably. | * bigphysarea causes system hang when driver calls it and "bigphysarea=NNNN" hasn't been given as a kernel option. This is bigphysarea's fault, probably. | ||
| + | * "EMPTY" command doesn't update the DSP; this is ok as long as acq_config is used for the next acquisition. It is broken when the user is issuing "go rc# ret_dat" manually. | ||
== mce_cmd / mce_library == | == mce_cmd / mce_library == | ||
| − | |||
* "cc fw_rev" expects at most 1 argument; this prevents us from doing, e.g. "rra cc fw_rev 0 1", which should be allowed. | * "cc fw_rev" expects at most 1 argument; this prevents us from doing, e.g. "rra cc fw_rev 0 1", which should be allowed. | ||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
== dsp_cmd / dsp_library == | == dsp_cmd / dsp_library == | ||
| Line 44: | Line 27: | ||
then tried "maslog reopen" and that didn't help either. | then tried "maslog reopen" and that didn't help either. | ||
| + | |||
| + | [[Category:MAS]] | ||
Latest revision as of 16:22, 30 August 2016
device driver
- dsp_ops: write_callback awakens the read sleeper before updating the command state. That leads to occasional dsp_cmd hangups.
- dsp_pci/dsp_driver: there are race conditions in the cleanup/remove procedure. Interrupts aren't disabled until after higher levels de-init, meaning callbacks get sent into nowhere.
- Driver doesn't work well on realtime (in NFY/HST mode). Oddly, non-realtime acquisitions at 400 Hz in this mode have been achieved.
- Inform interval should be remembered, and updated whenever necessary. [ Careful with this, you have to watch for things like DSP resets / mode switches! ]
- Incorrect DSP packet size leads to DSP dropping the packet and no one knows to time out.
- bigphysarea causes system hang when driver calls it and "bigphysarea=NNNN" hasn't been given as a kernel option. This is bigphysarea's fault, probably.
- "EMPTY" command doesn't update the DSP; this is ok as long as acq_config is used for the next acquisition. It is broken when the user is issuing "go rc# ret_dat" manually.
mce_cmd / mce_library
- "cc fw_rev" expects at most 1 argument; this prevents us from doing, e.g. "rra cc fw_rev 0 1", which should be allowed.
dsp_cmd / dsp_library
maslog_server
- when idl was run and set_directory set a new data directory, log file disappeared and there was a period where commands where not
logged in yesterday/log and there was no today/log either.
- 2007-12-03 MA: noticed that $MAS_DATA/log is empty and although commands go through nothing is logged. tried to reload mas.cfg:
maslog_server -f /etc/mas.cfg maslog_server - logging server for MAS load_config: key 'level' not found in config file listen_on_addr: failed to bind [98] (address already in use?) listener_listen: could not listen on address 'localhost:7894'
then tried "maslog reopen" and that didn't help either.
