Thread stacks.
Auto-unprotect not doing its thing
- Thread starter Black Hole
- Start date
MymsMan
Ad detector
I am not sure that it is fixed, I've just installed the latest Recmon, Epg and Webif packages. Stopped and restarted Recmon from services menu.
Still have a large and increasing number of anon entries
Code:
pmap `pgrep -f sbin/recmon`
15556: /mod/sbin/recmon -d -D /media/My Video
00400000 16K r-x-- /mnt/hd2/mod/sbin/recmon
00413000 4K rw--- /mnt/hd2/mod/sbin/recmon
00414000 76K rwx-- [ anon ]
2aaa8000 28K r-x-- /lib/ld-uClibc-0.9.29.so
2aaaf000 4K r---- /lib/ld-uClibc-0.9.29.so
2aab0000 4K rw--- /lib/ld-uClibc-0.9.29.so
2aab1000 8K rw--- [ anon ]
2aab3000 28K r-x-- /mnt/hd2/mod/lib/libinotifytools.so.0.4.1
2aaba000 4K rw--- /mnt/hd2/mod/lib/libinotifytools.so.0.4.1
2aabb000 76K rw--- [ anon ]
2aace000 64K r-x-- /lib/libpthread-0.9.29.so
2aade000 4K r---- /lib/libpthread-0.9.29.so
2aadf000 4K rw--- /lib/libpthread-0.9.29.so
2aae0000 8K rw--- [ anon ]
2aae2000 628K r-x-- /lib/libuClibc-0.9.29.so
2ab7f000 4K r---- /lib/libuClibc-0.9.29.so
2ab80000 4K rw--- /lib/libuClibc-0.9.29.so
2ab81000 24K rw--- [ anon ]
2ab87000 8K r-x-- /lib/libdl-0.9.29.so
2ab89000 4K r---- /lib/libdl-0.9.29.so
2ab8a000 4K rw--- /lib/libdl-0.9.29.so
2ab8b000 4K ----- [ anon ]
2ab8c000 1020K rw--- [ anon ]
2ac8b000 4K ----- [ anon ]
2ac8c000 1020K rw--- [ anon ]
2ad8b000 4K ----- [ anon ]
2ad8c000 1020K rw--- [ anon ]
2ae8b000 4K ----- [ anon ]
2ae8c000 1020K rw--- [ anon ]
2af8b000 4K ----- [ anon ]
2af8c000 1020K rw--- [ anon ]
2b08b000 4K ----- [ anon ]
2b08c000 1020K rw--- [ anon ]
2b18b000 4K ----- [ anon ]
2b18c000 1020K rw--- [ anon ]
2b28b000 4K ----- [ anon ]
2b28c000 1020K rw--- [ anon ]
2b38b000 4K ----- [ anon ]
2b38c000 1020K rw--- [ anon ]
2b48b000 4K ----- [ anon ]
2b48c000 1020K rw--- [ anon ]
2b58b000 4K ----- [ anon ]
2b58c000 1020K rw--- [ anon ]
2b68b000 4K ----- [ anon ]
2b68c000 1020K rw--- [ anon ]
2b78b000 4K ----- [ anon ]
2b78c000 1020K rw--- [ anon ]
2b88b000 4K ----- [ anon ]
2b88c000 1020K rw--- [ anon ]
2b98b000 4K ----- [ anon ]
2b98c000 1020K rw--- [ anon ]
2ba8b000 4K ----- [ anon ]
2ba8c000 1020K rw--- [ anon ]
2bb8b000 4K ----- [ anon ]
2bb8c000 1020K rw--- [ anon ]
7fcc5000 84K rwx-- [ stack ]
total 18496Ktotal 23632K
I have a small number but it isn't increasing, even when I force through multiple recmon actions.
Code:
humax# pmap 22455 | grep -c 1020K
15Has yours stabilised at all?
I think I need to drop the size of that thread stack though - 1MiB per thread is excessive for this.
MymsMan
Ad detector
The highest I reached was 34952K before I restarted Recmon again.
I have 8 entries in my recmon.d so after reboot with 9 stack frames the size is total 10272K
I will monitor usage over the day and see how it goes.
There could be significant bursts of close to simultaneous activity if you select a list of files for moving/deletion (especially with the FlatView clones),
DetectAds also create multiple working files when triggered for recording start and then there is the deletion and renaming of files when detection is completed.
Reducing stack frame size seems like a good idea - very little processing is done under the tasks spawned by recmon, the detectads task just starts, if needed, another task to run asynchronously to do the real work.
I have 8 entries in my recmon.d so after reboot with 9 stack frames the size is total 10272K
I will monitor usage over the day and see how it goes.
There could be significant bursts of close to simultaneous activity if you select a list of files for moving/deletion (especially with the FlatView clones),
DetectAds also create multiple working files when triggered for recording start and then there is the deletion and renaming of files when detection is completed.
Reducing stack frame size seems like a good idea - very little processing is done under the tasks spawned by recmon, the detectads task just starts, if needed, another task to run asynchronously to do the real work.
The new version will use fewer threads in general and a smaller stack size:
The stack only needs to be big enough for spawning the job - the job itself runs as a separate process.
Code:
2ab8c000 12K rw--- [ anon ]
2ab8f000 4K ----- [ anon ]
2ab90000 12K rw--- [ anon ]
2ab93000 4K ----- [ anon ]The stack only needs to be big enough for spawning the job - the job itself runs as a separate process.
MymsMan
Ad detector
I tried moving 33 files from the dustbin to My Video and the stack count increased from 9 to 25,
I then deleted the 33 files again and the stack count increased from 25 to 28,
I then started a recording with ad detection and the stack count increased from 28 to 40 (total 42276K)
Stopping the recording and starting another two recordings didn't change the number of stacks so hopefully I have reached the high water mark.
I will repeat the tests with the new version, when available, even if it uses the same number of threads 40x16K will be significantly less than 40x1MB
I then deleted the 33 files again and the stack count increased from 25 to 28,
I then started a recording with ad detection and the stack count increased from 28 to 40 (total 42276K)
Stopping the recording and starting another two recordings didn't change the number of stacks so hopefully I have reached the high water mark.
I will repeat the tests with the new version, when available, even if it uses the same number of threads 40x16K will be significantly less than 40x1MB
prpr
Well-Known Member
Isn't that mostly just an address space allocation anyway and not committed storage?
I'm sure it'll be enough for anyone.
New version now available. It limits thread concurrency and uses the smaller stack.
Yes but still better to keep it smaller. In the original version recmon was eventually unable to create a new thread as the old ones were never cleaned up.
MymsMan
Ad detector
Repeating most of my tests with the latest version I peaked at 26 threads and total 1548KI tried moving 33 files from the dustbin to My Video and the stack count increased from 9 to 25,
I then deleted the 33 files again and the stack count increased from 25 to 28,
I then started a recording with ad detection and the stack count increased from 28 to 40 (total 42276K)
Stopping the recording and starting another two recordings didn't change the number of stacks so hopefully I have reached the high water mark.
I will repeat the tests with the new version, when available, even if it uses the same number of threads 40x16K will be significantly less than 40x1MB
I think the chances of Recmon running out of allocatable storage are now highly remote!
MymsMan
Ad detector
Typically as soon as I turned off debugging logging for recmon it failed to generate a -start event for a recording and so DetectAds failed to start for that program. 
Recmon did generate other events at around the same time and there are no dumps so it was running. Start events have been generated for most recording but I am also missing another -start
I will turn on logging again and see if it shows any details but it reminds me of a problem we had a while back when recmon started to generate -start events
Recmon did generate other events at around the same time and there are no dumps so it was running. Start events have been generated for most recording but I am also missing another -start
I will turn on logging again and see if it shows any details but it reminds me of a problem we had a while back when recmon started to generate -start events