DPDK rte_hash multithreading - multithreading

everyone! I'm writing DPDK based application. I'm reading packets from NIC, enqueing them to the ring. Then I have multiple worker threads which dequeues packets from rx ring, parse their headers to get destination ip address and level 4 protocol destination port. This data is packed into structure:
struct session_key {
rte_be32_t ip_dst;
rte_be16_t port_dst;
};
This structure is used as a key in rte_hash table. As data in this hash table I use uint32_t counter which increments when packet matches key. I create rte_hash with RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY flag to make it thread safe for multithreaded read and write.
Each worker thread gets dst_ip and dst_port from packet and lookups hash table for such key. If key exists, its value incremented, if key does not exists it is added to the table with data = 1.
uint32_t *found;
int ret = rte_hash_lookup_data(sessions_hash_table, (void *)&key, (void **)&found);
if (ret < 0) {
uint32_t *data = rte_zmalloc("session_key", sizeof(uint32_t), 0);
*data = 1;
rte_hash_add_key_data(sessions_hash_table, &key, data);
} else {
(*found)++;
}
So I have multiple readers, writers to the hash table. After all workers are finished common statistics is calculated in main thread. How much packets match each pair of ip and port is printed on screen.
The problem is that when I use only one worker there is no problem, amount of received packets is equal to packets saved in hash table.
But when I use several worker threads I get unequal numbers. I understand that there is probability of reading table while another thread is writing to it. But I thought that config flags like RTE_HASH_EXTRA_FLAGS_RW_CONCURRENCY will help me to deal with multithreading.
So I need some advice on how to make rte_hash work in multithreaded application where several threads write and read the same hash table.

Related

netfilter hook is not retrieving complete packet

I'm writing a netfilter module, that deeply inspect the packet. However, during tests I found that netfilter module is not receiving the packet in full.
To verify this, I wrote the following code to dump packet retrieved on port 80 and write the result to dmesg buffer:
const struct iphdr *ip_header = ip_hdr(skb);
if (ip_header->protocol == IPPROTO_TCP)
{
const struct tcphdr *tcp_header = tcp_hdr(skb);
if (ntohs(tcp_header->dest) != 80)
{
return NF_ACCEPT;
}
buff = (char *)kzalloc(skb->len * 10, GFP_KERNEL);
if (buff != NULL)
{
int pos = 0, i = 0;
for (i = 0; i < skb->len; i ++)
{
pos += sprintf(buff + pos, "%02X", skb->data[i] & 0xFF);
}
pr_info("(%pI4):%d --> (%pI4):%d, len=%d, data=%s\n",
&ip_header->saddr,
ntohs(tcp_header->source),
&ip_header->daddr,
ntohs(tcp_header->dest),
skb->len,
buff
);
kfree (buff);
}
}
In virtual machine running locally, I can retrieve the full HTTP request; On Alibaba cloud, and some other OpenStack based VPS provider, the packet is cut in the middle.
To verify this, I execute curl http://VPS_IP on another VPS, and I got the following output in dmesg buffer:
[ 1163.370483] (XXXX):5007 --> (XXXX):80, len=237, data=451600ED000040003106E3983D87A950AC11D273138F00505A468086B44CE19E80180804269300000101080A1D07500A000D2D90474554202F20485454502F312E310D0A486F73743A2033392E3130372E32342E37370D0A4163636570743A202A2F2A0D0A557365722D4167656E743A204D012000000000000000000000000000000000000000000000000000000000000000000000010000000000000000000000001E798090F5FFFF8C0000007B00000000E0678090F5FFFF823000003E00000040AE798090F5FFFF8C0000003E000000000000000000000000000000000000000000000000000000000000
When decoded, the result is like this
It's totally weird, everything after User-Agent: M is "gone" or zero-ed. Although the skb->len is 237, but half of the packet is missing.
Any ideas? Tried both PRE_ROUTING and LOCAL_IN, no changes.
It appears that sometimes you are getting a linear skb, and sometimes your skb is not linear. In the latter case you are not reading the full data contents of an skb.
If skb->data_len is zero, then your skb is linear and the full data contents of the skb is in skb->data. If skb->data_len is not zero, then your skb is not linear, and skb->data contains just the the first (linear) part of the data. The length of this area is skb->len - skb->data_len. skb_headlen() helper function calculates that for convenience. skb_is_nonlinear() helper function tells in an skb is linear or not.
The rest of the data can be in paged fragments, and in skb fragments, in this order.
skb_shinfo(skb)->nr_frags tells the number of paged fragments. Each paged fragment is described by a data structure in the array of structures skb_shinfo(skb)->frags[0..skb_shinfo(skb)->nr_frags]. skb_frag_size() and skb_frag_address() helper functions help dealing with this data. They accept the address of the structure that describes a paged fragment. There are other useful helper functions depending on your kernel version.
If the total size of data in paged fragments is less than skb->data_len, then the rest of the data is in skb fragments. It's the list of skb which is attached to this skb at skb_shinfo(skb)->frag_list (see skb_walk_frags() in the kernel).
Please note that there may be that there's no data in the linear part and/or there's no data in the paged fragments. You just need to process data piece by piece in the order just described.

ping pong DMA in kernel

I am using a SoC to DMA data from the programmable logic side (PL) to the processor side (PS). My overall scheme is to allocate two separate dma buffers and ping-pong data to these buffers to the user-space application can access data without collisions. I have successfully done this using single dma transactions in a loop (in a kthread) with each transaction either on the first or second buffer. I was able to notify a poll() method of either the first or second buffer.
Now I would like to investigate scatter-gather and/or cyclic DMA.
struct scatterlist sg[2]
struct dma_async_tx_descriptor *chan_desc;
struct dma_slave_config config;
sg_init_table(sg, ARRAY_SIZE(sg));
addr_0 = (unsigned int *)dma_alloc_coherent(NULL, dma_size, &handle_0, GFP_KERNEL);
addr_1 = (unsigned int *)dma_alloc_coherent(NULL, dma_size, &handle_1, GFP_KERNEL);
sg_dma_address(&sg[0]) = handle_0;
sg_dma_len(&sg[0]) = dma_size;
sg_dma_address(&sg[1]) = handle_1;
sg_dma_len(&sg[1]) = dma_size;
memset(&config, 0, sizeof(config));
config.direction = DMA_DEV_TO_MEM;
config.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
config.src_maxburst = DMA_MAX_BURST_SIZE / DMA_SLAVE_BUSWIDTH_4_BYTES;
dmaengine_slave_config(dma_dev->chan, &config);
chan_desc = dmaengine_prep_slave_sg(dma_dev->chan, &sg[0], 2, DMA_DEV_TO_MEM, DMA_CTRL_ACK | DMA_PREP_INTERRUPT);
chan_desc->callback = sync_callback;
chan_desc->callback_param = dma_dev->cmp;
init_completion(dma_dev->cmp);
dmaengine_submit(chan_desc);
dma_async_issue_pending(dma_dev->chan);
dma_free_coherent(NULL, dma_size, addr_0, handle_0);
dma_free_coherent(NULL, dma_size, addr_1, handle_1);
This works fine for a single run through the scatterlist and then calls the call_back at sync_callback. My thought was to chain the scatterlist in a loop but I won't get the callback.
IS THERE A WAY to have a callback for each descriptor in the scatterlist? I wondered if I could use dmaengine_prep_slave_cyclic (calls callback after every transaction) but it looks to me like this is for a single buffer when reviewing dmaengine.h. Looking at DMA Engine API Guide it looks like there is another option dmaengine_prep_interleaved_dma using a dma_interleaved_template that sounds interesting but hard to find info about.
In the end I just want to signal the user space in some manner as to what buffer is ready.

why it's slowly when I parse a message of Google protocol buffer in multi-thread?

I try to parse many Google protocol buffer messages from a binary file generated by calling SerializeToString. I first load all Bytes into a heap memory by calling new function. I also have two arrays to store the Bytes begin address of a message in the heap memory and the Bytes count of the message.
Then I begin to parse message by calling ParseFromString.I want to quicken the procedure by using multi-thread.
In each thread, I pass the start index and end index of address array and Byte count array.
In parent process. the main code is:
struct ParsePara
{
char* str_buffer;
size_t* buffer_offset;
size_t* binary_string_length_array;
size_t start_idx;
size_t end_idx;
Flight_Ticket_Info* ticket_info_buffer_array;
};
//Flight_Ticket_Info is class of message
//offset_size is the count of message
ticket_array = new Flight_Ticket_Info[offset_size];
const int max_thread_count = 6;
pthread_t pthread_id_vec[max_thread_count];
CTimer thread_cost;
thread_cost.start();
vector<ParsePara*> para_vec;
const size_t each_count = ceil(float(offset_size) / max_thread_count);
for (size_t k = 0;k < max_thread_count;k++)
{
size_t start_idx = each_count * k;
size_t end_idx = each_count * (k+1);
if (start_idx >= offset_size)
break;
if (end_idx >= offset_size)
end_idx = offset_size;
ParsePara* cand_para_ptr = new ParsePara();
if (!cand_para_ptr)
{
_ERROR_EXIT(0,"[Malloc memory fail.]");
}
cand_para_ptr->str_buffer = m_valdata;//heap memory for storing Bytes of message
cand_para_ptr->buffer_offset = offset_array;//begin address of each message
cand_para_ptr->start_idx = start_idx;
cand_para_ptr->end_idx = end_idx;
cand_para_ptr->ticket_info_buffer_array = ticket_array;//array to store message
cand_para_ptr->binary_string_length_array = binary_length_array;//Bytes count of each message
para_vec.push_back(cand_para_ptr);
}
for(size_t k = 0 ;k < para_vec.size();k++)
{
int ret = pthread_create(&pthread_id_vec[k],NULL,parserFlightTicketForMultiThread,para_vec[k]);
if (0 != ret)
{
_ERROR_EXIT(0,"[Error] [create thread fail]");
}
}
for (size_t k = 0;k < para_vec.size();k++)
{
pthread_join(pthread_id_vec[k],NULL);
}
In each thread the thread function is:
void* parserFlightTicketForMultiThread(void* void_para_ptr)
{
ParsePara* para_ptr = (ParsePara*) void_para_ptr;
parserFlightTicketForMany(para_ptr->str_buffer,para_ptr->ticket_info_buffer_array,para_ptr->buffer_offset,
para_ptr->start_idx,para_ptr->end_idx,para_ptr->binary_string_length_array);
}
void parserFlightTicketForMany(const char* str_buffer,Flight_Ticket_Info* ticket_info_buffer_array,
size_t* buffer_offset,const size_t start_idx,const size_t end_idx,size_t* binary_string_length_array)
{
printf("start_idx:%d,end_idx:%d\n",start_idx,end_idx);
for (size_t k = start_idx;k < end_idx;k++)
{
if (k % 100000 == 0)
cout << k << endl;
size_t cand_offset = buffer_offset[k];
size_t binary_length = binary_string_length_array[k];
ticket_info_buffer_array[k].ParseFromString(string(&str_buffer[cand_offset],binary_length-1));
}
printf("done %ld %ld\n",start_idx,end_idx);
}
But multi-thread cost is more than one thread.
one thread cost is:40455623ms
My computer is 8 core and six thread cost is:131586865ms
Anyone can help me? thank you!
Some possible problems -- you'll have to experiment to determine which:
Protobuf parsing speed is often limited by memory bandwidth rather than CPU time, especially with a large input data set. In that case, more threads won't help, since all the cores are sharing bandwidth to main memory. Indeed, having multiple cores fighting over memory bandwidth could make the overall operation slower. Note that the biggest consumer of memory is not the input bytes but rather the parsed data objects -- that is, the output of parsing -- which are many times larger than the encoded data. To improve this problem, consider writing the parsing loop so that it fully-processes each message immediately after parsing, before moving on to the text message. That way, instead of allocating k protobuf objects, you only need to allocate one protobuf object per thread, and repeatedly reuse the same object for parsing. This way the object will (probably) stay in the core's private L1 cache and avoid consuming memory bandwidth; only the input bytes will be read over the main bus.
How are you loading data into RAM? Did you read() into a large array or did you mmap()? In the latter case the data is read from disk lazily -- it won't happen until you actually attempt to parse it. Even in the read() case, it could be that the data has been swapped out, creating similar effects. Either way, your threads are now not just fighting for memory bandwidth, but disk bandwidth, which is of course much slower. Having six threads reading separate parts of a big file will definitely be slower overall than having one thread read the whole file, because the operating system optimizes for sequential access.
Protobuf allocates memory during parsing. Many memory allocators take a lock while allocating new memory. Since all your threads are allocating tons and tons of objects in a tight loop, they will contend for this lock. Make sure you are using a thread-friendly memory allocator, such as Google's tcmalloc. Note that repeatedly reusing the same protobuf object in a parse-consume loop rather than allocating lots of different objects will also help immensely here, because the protobuf object will automatically reuse memory for sub-objects.
There may be a bug in your code and it might not be doing what you expect at all when multithreaded. For example, a bug might be causing all the threads to process the same data, rather than different data, and it could be that the data they're choosing happens to be bigger. Make sure you are testing that the results of your code are exactly the same when you run single-threaded vs. multi-threaded.
In short, if you want multiple cores to make your code faster, you have to think about not just what each core is doing, but what data is going in and out of each core, and how much the cores have to talk to each other. Ideally you want each core to operate all on its own without talking to anyone or anything; then you get maximum parallelism. That's not usually possible, of course, but the closer you can get to that, the better.
BTW, a random optimization for you:
ParseFromString(string(&str_buffer[cand_offset],binary_length-1))
Replace that with:
ParseFromArray(&str_buffer[cand_offset],binary_length-1)
Creating at std::string makes a copy of the data, which wastes time (and memory bandwidth). (This doesn't explain why threading is slow, though.)

recv with flags MSG_DONTWAIT | MSG_PEEK on TCP socket

I have a TCP stream connection used to exchange messages. This is inside Linux kernel. The consumer thread keeps processing incoming messages. After consuming one message, I want to check if there are more pending messages; in which case I would process them too. My code to achieve this looks like below. krecv is wrapper for sock_recvmsg(), passing value of flags without modification (krecv from ksocket kernel module)
With MSG_DONTWAIT, I am expecting it should not block, but apparently it blocks. With MSG_PEEK, if there is no data to be read, it should just return zero. Is this understanding correct ? Is there a better way to achieve what I need here ? I am guessing this should be a common requirement as message passing across nodes is used frequently.
int recvd = 0;
do {
recvd += krecv(*sockp, (uchar*)msg + recvd, sizeof(my_msg) - recvd, 0);
printk("recvd = %d / %lu\n", recvd, sizeof(my_msg));
} while(recvd < sizeof(my_msg));
BUG_ON(recvd != sizeof(my_msg));
/* For some reason, below line _blocks_ even with no blocking flags */
recvd = krecv(*sockp, (uchar*)tempbuf, sizeof(tempbuf), MSG_PEEK | MSG_DONTWAIT);
if (recvd) {
printk("more data waiting to be read");
more_to_process = true;
} else {
printk("NO more data waiting to be read");
}
You might check buffer's length first :
int bytesAv = 0;
ioctl(m_Socket,FIONREAD,&bytesAv); //m_Socket is the socket client's fd
If there are data in it , then recv with MSG_PEEK should not be blocked ,
If there are no data at all , then no need to MSG_PEEK ,
that might be what you like to do .
This is a very-very old question, but
1. problem persits
2. I faced with it.
At least for me (Ubuntu 19.04 with python 2.7) this MSG_DONTWAIT has no effect, however if I set the timeout to zero (with settimeout function), it works nicely.
This can be done in c with setsockopt function.

Looking for a lock-free RT-safe single-reader single-writer structure

I'm looking for a lock-free design conforming to these requisites:
a single writer writes into a structure and a single reader reads from this structure (this structure exists already and is safe for simultaneous read/write)
but at some time, the structure needs to be changed by the writer, which then initialises, switches and writes into a new structure (of the same type but with new content)
and at the next time the reader reads, it switches to this new structure (if the writer multiply switches to a new lock-free structure, the reader discards these structures, ignoring their data).
The structures must be reused, i.e. no heap memory allocation/free is allowed during write/read/switch operation, for RT purposes.
I have currently implemented a ringbuffer containing multiple instances of these structures; but this implementation suffers from the fact that when the writer has used all the structures present in the ringbuffer, there is no more place to change from structure... But the rest of the ringbuffer contains some data which don't have to be read by the reader but can't be re-used by the writer. As a consequence, the ringbuffer does not fit this purpose.
Any idea (name or pseudo-implementation) of a lock-free design? Thanks for having considered this problem.
Here's one. The keys are that there are three buffers and the reader reserves the buffer it is reading from. The writer writes to one of the other two buffers. The risk of collision is minimal. Plus, this expands. Just make your member arrays one element longer than the number of readers plus the number of writers.
class RingBuffer
{
RingBuffer():lastFullWrite(0)
{
//Initialize the elements of dataBeingRead to false
for(unsigned int i=0; i<DATA_COUNT; i++)
{
dataBeingRead[i] = false;
}
}
Data read()
{
// You may want to check to make sure write has been called once here
// to prevent read from grabbing junk data. Else, initialize the elements
// of dataArray to something valid.
unsigned int indexToRead = lastFullWriteIndex;
Data dataCopy;
dataBeingRead[indexToRead] = true;
dataCopy = dataArray[indexToRead];
dataBeingRead[indexToRead] = false;
return dataCopy;
}
void write( const Data& dataArg )
{
unsigned int writeIndex(0);
//Search for an unused piece of data.
// It's O(n), but plenty fast enough for small arrays.
while( true == dataBeingRead[writeIndex] && writeIndex < DATA_COUNT )
{
writeIndex++;
}
dataArray[writeIndex] = dataArg;
lastFullWrite = &dataArray[writeIndex];
}
private:
static const unsigned int DATA_COUNT;
unsigned int lastFullWrite;
Data dataArray[DATA_COUNT];
bool dataBeingRead[DATA_COUNT];
};
Note: The way it's written here, there are two copies to read your data. If you pass your data out of the read function through a reference argument, you can cut that down to one copy.
You're on the right track.
Lock free communication of fixed messages between threads/processes/processors
fixed size ring buffers can be used in lock free communications between threads, processes or processors if there is one producer and one consumer. Some checks to perform:
head variable is written only by producer (as an atomic action after writing)
tail variable is written only by consumer (as an atomic action after reading)
Pitfall: introduction of a size variable or buffer full/empty flag; these are typically written by both producer and consumer and hence will give you an issue.
I generally use ring buffers for this purpoee. Most important lesson I've learned is that a ring buffer of can never contain more than elements. This way a head and tail variable are written by producer respectively consumer.
Extension for large/variable size blocks
To use buffers in a real time environment, you can either use memory pools (often available in optimized form in real time operating systems) or decouple allocation from usage. The latter fits to the question, I believe.
If you need to exchange large blocks, I suggest to use a pool with buffer blocks and communicate pointers to buffers using a queue. So use a 3rd queue with buffer pointers. This way the allocates can be done in application (background) and you real time portion has access to a variable amount of memory.
Application
while (blockQueue.full != true)
{
buf = allocate block of memory from heap or buffer pool
msg = { .... , buf };
blockQueue.Put(msg)
}
Producer:
pBuf = blockQueue.Get()
pQueue.Put()
Consumer
if (pQueue.Empty == false)
{
msg=pQueue.Get()
// use info in msg, with buf pointer
// optionally indicate that buf is no longer used
}

Resources