we've had some troubles with TWI/I2C after waking up from sleep with the Atmel Xmega256A3. Instead of digging into the details of TWI/I2C we've decided to use the supplied twi_master_driver from Atmel attached to AVR1308 application note.
The problem is one or a few failed TWI transactions just after waking up from sleep. On the I2C-bus connected to the XMega we have a few potentiometers, a thermometer and an RTC. The XMega acts as the only master on the bus.
We use the sleep functions found in AVRLIBC:
{code for turning of VCC to all I2C connected devices}
set_sleep_mode(SLEEP_MODE_PWR_DOWN);
sleep_enable();
sleep_cpu();
{code for turning on VCC to all I2C connected devices}
The XMega as woken from sleep by the RTC which sets a pin high. After the XMega is woken from sleep, we want to set a value on one of the potentiometers, but this fails. For some reason, the TWI-transaction result is TWIM_RESULT_NACK_RECEIVED instead of TWIM_RESULT_OK in the first transaction. After that everything seems to work again.
Have we missed anything here? Is there any known issues with the XMega, sleep and TWI? Do we need to reset the TWI of clear any flags after waking from sleep?
Best regards
Fredrik
There is a common problem on I2C/TWI where the internal state machine gets stuck in an intermediate state if a transaction is not completed fully. The slave then does not respond correctly when addressed on the next transaction. This commonly happens when the master is reset or stops outputting the SCK signal part way through the read or write. A solution is to toggle the SCK line manually 8 or 9 times before starting any data transactions so the that the internal state machines in the slaves are all reset to the start of transfer point and they are all then looking for their address byte.
Related
I'm trying to track down a stall that may or may not be on the client host, but on the server side. Unfortunately this is all kernel RPC level, with some of if not controlled by my code...
I'm something of a neophyte when it comes to rpc debugging, so any references would be helpful.
I have this output in dmesg when rpcdebug -c -m all is run:
[2109401.599881] -pid- flgs status -client- --rqstp- -timeout ---ops--
[2109401.600055] 51580 0880 0 ffff9af4c4da9800 ffff9af4c4416600 15000 ffffffffc0b26680 nfsv3 GETATTR a:call_status [sunrpc] q:xprt_pending
[2109401.600300] 51581 0880 0 ffff9af4c4da9800 ffff9af42465f800 15000 ffffffffc0b26680 nfsv3 GETATTR a:call_status [sunrpc] q:xprt_pending
I get the PID, flags, and timeout, but:
What are the "client" and "rqstp" fields supposed to mean? If I see either duplicated in subsequent outputs, does that mean the RPC is stalled? And which way?
Is "xprt_pending" a "waiting to send the RPC" queue? If that queue was "delayq," I would know what that means in the context of the problem we're trying to diagnose. But this state doesn't seem to be explained anywhere I find in Google. (And my Google-Fu is usually better than this...)
What is the "ffffffffc0b26680" supposed to be? It repeats all over the output, for EVERY RPC listed.
I'm trying to avoid running with rpcdebug set, because I'm dealing with an intermittent stall, and I'd rather not slow EVERYTHING down in the hopes of catching the stall.
I am using veins4.6 with sumo 0.30 and omnet++5.1.1 in ubuntu 14.04. I have created a custom network with a cross(one intersection with 4 roads) and ran the simulation with 200 vehicles. I did not observe this behaviour for 4vehicles. I have seen it with 50 vehicles too. I need to get the count of total lost packets for my masters project. So I was looking at statistics and found that RXTXLostPackets is not zero. As far as I understood from documentation it should be zero if allowTxDuringRx=false. Default is false(PhyLayer80211p.ned). As I did not change any code yet, I was confused if that is expected behaviour.
What I have done so far.
from Mac1609_4::handleLowerControl, statsTXRXLostPackets is updated when Decider80211p responds with RECWHILESEND.
In Decider80211p::processSignalEnd, if value of whileSending is true RECWHILESEND is sent to mac layer as control message.
In Decider80211p::processSignalEnd, if(frame->getWasTransmitting() || phy11p->getRadioState() == Radio::TX) , this frame was considered as received while sending and sets the value for whileSending as true.
The wasTransmitting varilable is set to true in Decider80211p::switchToTx and Decider80211p::processNewSignal functions.
currentFrame->setWasTransmitting(true);
currentFrame->setBitError(true);
in Decider80211p::processNewSignal:
if (phy11p->getRadioState() == Radio::TX ) {
frame->setBitError(true); --> tried disabling both these values and the RXTXLostPackets was zero.
frame->setWasTransmitting(true);
DBG_D11P << "AirFrame: " << frame->getId() << " (" << recvPower << ") received, while already sending. Setting BitErrors to true" << std::endl;
}
There is one thread with similar issue with the fix of adding this line in processSignalEnd function. But looks like veins4.6 does not use curSyncFrame anymore.
Veins - Unexpected behavior with lost packets in certain vehicles
if (!frame->getWasTransmitting()){
curSyncFrame = 0;
}
I could not clearly understand the issue. The code and configuration files I have used are here. https://github.com/Rajeswar59/veins_learning.
Can anyone please take a look and help me with this. Thanks in advance.
edit: I went through the logs. This is what I could understand as of now.
https://drive.google.com/open?id=0BzjDW8PQhkSmSEUtZ2lpcld4ZXc --> some portion of logs are here.
---> order of sending
#13332 0.247987176594 node[30] --> node[48] id=22266
#13375 0.247987796864 node[18] --> node[20] id=22447
#13384 0.247987864534 node[20] --> node[30] id=22573
From logs I have concentrated on node 18. Two nodes that transmitted before 30 are 32 and 4. These 2 messages are received successfully by all 3 nodes. When a message arrives decider tries to set channel state as busy in processnewsignal and set idle after processing packet. This calls mac1609_4.cc channelBusy and channelIdle functions respectively. So the channelIdle variable is set accordingly. Also if channel is to be set busy it will stop contention and calculate currentBackoff if any packet is waiting to be transmitted. If channel is being set idle at the end of reception, startContent is called. Based on this only the lastIdle variable is set which is used to calculate nextMacEvent. So when the last successful message was received all the nodes which have a packet to send decide nextMacEvent and it is sent as self message in Mac1609_4.cc. on receiving the nextMacEvent self message we will start transmitting without checking if any other node has started transmission. We can not identify that probably because we are setting channel busy when we receive messages after some propagation delay. So between last successful transmission and nextMacEvent other nodes also take decision to transmit without checking current channel state. That's why the node has some receive events while sending. As far as my understanding goes before transmission we should sense current state of channel and retry backoff accordingly. We do not check this at the nextMacEvent. It looks like a collision behaviour but should we not check the current state of channel when backoff counter reaches zero and retry. Please correct me if I am wrong anywhere.
Thanks for your patience.
Any help or advice??
My Learnings(probably last update):
After Some digging, these are my learnings if it helps some one. The basic CSMA mechanism says before attempting for transmission, the node has to sense the channel, initiate transmission if the channel is sensed idle for AIFS time, or go in to back off if channel is busy. In veins the channel busy status is stored in idleChannel variable whose status is checked in Mac1609_4:channelBusySelf() function before initiating transmission (nextMacEvent in Mac1609_4::handleSelfMsg). The idleChannel is updated in Mac1609_4::channelBusy and Mac1609_4::channelIdle functions when a message reception starts and when message reception ends respectively. So when a previously transmitting node sends a packet, all the recieving nodes will receive the packet with varying delay i.e., starts receiving at different times and update their channelIdle variable. After that they calculate best time to transmit and starts transmission. It does check if channel is idle or not but as the channelIdle status is updated at next reception and because of transmission delay it takes some time between transmission start at sender and reception start at receiver side, both the transmitting nodes cant see other transmission. As far as I understand this is called a collision when more than two nodes start transmission at the same time. So the BitError statistic is set and statsTXRXLostPackets is also set. So while calculating totalLostPackets we can take only one of these two values.
I tried to solve the problem, but I got a different table than the table that xilinx shows. I attatched both my answer and real answer. Xilinx shows that "out" is 'U' until 36ns, after 36ns, it is '1'. Can anyone help me about why the "out" graphics is not assigned any value before 36ns?(I think it should be assigned first at 20 ns).
my answer
question
This turned out to be a really good question. I initially thought you had done something wrong when simulating, but then I ran my own simulation and got the same result.
It turns out that the a <= b after x assignment uses something called the "inertial time model" by default. In this mode scheduled events will be cancelled if b changes again before x time has passed. The purpose is to filter pulses shorter than the specified delay. In your case this is what the simulator will do:
At t=0, out is scheduled to change to 1 at t=20.
At t=12, tem1 or tem2 changes to 0. The scheduled change at t=20 is cancelled and a new change to 0 is scheduled at t=32.
At t=16, tem1 or tem2 changes back to 1. Again the scheduled change is cancelled and a new change is scheduled at t=36.
After this tem1 or tem2 remains at 1, so the change at t=36 is executed and out finally changes from U.
You can change to the "transport delay model" using out <= transport tem1 or tem2 after 20 ns; In this case your drawn waveform will match with simulation.
I have been testing MongoDB 2.6.7 for the last couple of months using YCSB 0.1.4. I have captured good data comparing SSD to HDD and am producing engineering reports.
After my testing was completed, I wanted to explore the allanbank async driver. When I got it up and running (I am not a developer, so it was a challenge for me), I first wanted to try the rebuilt sync driver. I found performance improvements of 30-100%, depending on the workload, and was very happy with it.
Next, I tried the async driver. I was not able to see much difference between it and my results with the native driver.
The command I'm running is:
./bin/ycsb run mongodb -s -P workloads/workloadb -p mongodb.url=mongodb://192.168.0.13:27017/ycsb -p mongodb.writeConcern=strict -threads 96
Over the course of my testing (mostly with the native driver), I have experimented with more and less threads than 96; turned on "noatime"; tried both xfs and ext4; disabled hyperthreading; disabled half my 12 cores; put the journal on a different drive; changed sync from 60 seconds to 1 second; and checked the network bandwidth between the client and server to ensure its not oversubscribed (10GbE).
Any feedback or suggestions welcome.
The Async move exceeded my expectations. My experience is with the Python Sync (pymongo) and Async driver (motor) and the Async driver achieved greater than 10x the throughput. further, motor is still using pymongo under the hoods but adds the async ability. that could easily be the case with your allanbank driver.
Often the dramatic changes come from threading policies and OS configurations.
Async needn't and shouldn't use any more threads than cores on the VM or machine. For example, if you're server code is spawning a new thread per incoming conn -- then all bets are off. start by looking at the way the driver is being utilized. A 4 core machine uses <= 4 incoming threads.
On the OS level, you may have to fine-tune parameters like net.core.somaxconn, net.core.netdev_max_backlog, sys.fs.file_max, /etc/security/limits.conf nofile and the best place to start is looking at nginx related performance guides including this one. nginx is the server that spearheaded or at least caught the attention of many linux sysadmin enthusiasts. Contrary to popular lore one should reduce your keepalive timeout opposed to lengthen it. The default keep-alive timeout is some absurd (4 hours) number of seconds. you might want to cut the cord in 1 minute. basically, think a short sweet relationship with your clients connections.
Bear in mind that Mongo is not Async so you can use a Mongo driver pool. nevertheless, don't let the driver get stalled on slow queries. cut it off in 5 to 10 seconds using the following equivalents in Java. I'm just cutting and pasting here with no recommendations.
# Specifies a time limit for a query operation. If the specified time is exceeded, the operation will be aborted and ExecutionTimeout is raised. If max_time_ms is None no limit is applied.
# Raises TypeError if max_time_ms is not an integer or None. Raises InvalidOperation if this Cursor has already been used.
CONN_MAX_TIME_MS = None
# socketTimeoutMS: (integer) How long (in milliseconds) a send or receive on a socket can take before timing out. Defaults to None (no timeout).
CLIENT_SOCKET_TIMEOUT_MS=None
# connectTimeoutMS: (integer) How long (in milliseconds) a connection can take to be opened before timing out. Defaults to 20000.
CLIENT_CONNECT_TIMEOUT_MS=20000
# waitQueueTimeoutMS: (integer) How long (in milliseconds) a thread will wait for a socket from the pool if the pool has no free sockets. Defaults to None (no timeout).
CLIENT_WAIT_QUEUE_TIMEOUT_MS=None
# waitQueueMultiple: (integer) Multiplied by max_pool_size to give the number of threads allowed to wait for a socket at one time. Defaults to None (no waiters).
CLIENT_WAIT_QUEUE_MULTIPLY=None
Hopefully you will have the same success. I was ready to bail on Python prior to async
Using Andrew Rapp's XBee-API, how can I sample I/O data via a coordinator from more than two endpoints?
I have 17 Series 1 XBees. I have programmed one to be a coordinator (API mode = 2) and the rest to be endpoints. Using XBee-API I am sending a Force I/O Sample ("IS") remote AT command, unicast to each endpoint. This works perfectly well when there are up to two endpoints, but as soon as a third is added, one of the three always becomes non-responsive (times out with XBeeTimeoutException). It's not always the same physical unit that stops responding, but it is always the third one (for example, if I send Force I/O Sample to Device1, Device2, and Device3, Device3 will time out, and if I change the order to Device3, Device1, Device2, Device2 will time out.
If I set up more than three XBees, about 1 out of 3 will time out - but not every third one.
I've verified that the XBees themselves are fine. I've searched the Internet and Stack Overflow in particular to no avail. I've tried using a simple ZNetRemoteAtRequest. I've tried opening and closing the XBee coordinator serial connection once for all three devices, once per device, and once per program run. I've tried varying the distance between the coordinator and endpoints (never more than five feet apart). I've tried different coordinator configuration parameters (from the Digi documentation). I've tried changing out the XBee for the coordinator.
This is the code I'm using to send the Force I/O Sample request to each endpoint and read the response:
xbee = new XBee(); // Coordinator
xbee.open("/dev/ttyUSB0, 115200)); // Happens before any of the endpoints are contacted
... // Loop through known endpoint addresses
XBeeRequest request = new ZBForceSampleRequest(new XBeeAddress64(endpointAddress));
ZNetRemoteAtResponse response = null;
response = (ZNetRemoteAtResponse) xbee.sendSynchronous(request, remoteXBeeTimeout);
if (response.isOk()) {
// Process response payload
}
... // End loop and finally close coordinator connection
What might help polling I/O samples from more than two endpoints?
EDIT: I found that Andrew Rapp's XBee-API library fakes multithreaded behavior, which causes the synchronization issues described in this question. I wrote a replacement library that is actually multithreaded and correctly maps responses from multiple XBee endpoints: https://github.com/steveperkins/xbee-api-for-java-1-4. When I wrote it Java 1.4 was necessary for use on the BeagleBone, Plug, and Zotac single-board PCs but it's an easy conversion to 1.7+.
Are you using hardware flow control on your serial port? Is it possible that you're sending requests out when the local XBee has deasserted CTS (e.g., asking you to stop sending)? I assume you're running at 115200 bps, so the XBee serial port can keep up with the network data rate.
Can you turn on debugging information, or connect some port monitoring hardware/software to display the data going over the serial port to the local XBee?