| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The way it's used now (upgrade to multislot for DL TBF upon LLC frame arrival)
doesn't seem right because DL and UL TBF do not exist at that moment.
Nevertheless, encoding routine is complete for EGPRS (no GPRS yet) and
trigger functions and state changes provide nice illustration into internal
workings of the scheduler. Hence we'll keep it here for the time being.
Signed-off-by: Max <msuraev@sysmocom.de>
|
|
Currently single BSN blocks of size 68 (a single unit from MCS-8)
are put into a MCS-6 block with padding according to TS 44.060
Annex J. Unfortunately, these packets are ignored by the E71. This
might be related to the currently incorrect usage of the puncturing
schemes and the RSB bit (the puncturing schemes are not updated and
the RBS flag is never set).
Avoid downgrading from MCS-8 to MCS-6 if there is no second block
available. Send the block twice instead.
Sponsored-by: On-Waves ehf
|
|
Comparing the coding scheme properties doesn't seem to be strong
enough, since create_dl_acked_block(fn, ts, index, index2) sometimes
had to combine data units of different size this way.
Check the registered size of the BSN blocks instead, since these must
match if two blocks are to be combined in a single RLC data message
with MCS7-9.
It is not yet clear, what exactly goes wrong with the current
implementation, but this commit fixes the problem which trigger the
assertion in create_dl_acked_block.
Sponsored-by: On-Waves ehf
|
|
Currently only one BSN block is encoded in each RLC data message,
even if MSC7-9 are used, which transport two independant (except for a
max BSN delta of 512) BSN blocks. In that case, the same block is
just put twice into the same message.
The current create_dl_acked_block(fn, ts) method handles
block selection (resend, new BSN, dummy block, ...) and restart
handling in one method and is too complex to extend it accordingly.
Therefore this commit move the block selection/creation handling into
a new method (take_next_bsn) which delivers the next BSN along with a
hint, whether it may be combined with another block. In that case,
the function can be called a second time (this time with a valid
previous BSN, that's the one returned by the first call) to get the
second BSN. The real block generation method
create_dl_acked_block(fn, ts, index, index2) is then called with both
BSNs as indices (the second must be -1, if there is only one BSN).
Note that every BSN returned by take_next_bsn should be passed to
create_dl_acked_block to avoid state inconsistencies.
Sponsored-by: On-Waves ehf
|
|
Currently the method is hard-coded to support a single BSN only.
MCS-7 to MCS-9 encode 2 data blocks into a single RLC data message.
This commit refactors create_dl_acked_block to process any number
of blocks internally.
Note that this does not extend the parameter list accordingly and
just duplicates the BSN if these MCS are being used.
Sponsored-by: On-Waves ehf
|
|
The offsets of the data areas change when padding is used (see TS
44.060, 9.3.2.1 and Annex J for details).
Extend the parameter lists to pass the with_padding flag and use
that information to compute the correct offsets.
Sponsored-by: On-Waves ehf
|
|
Currently only header type 3 (MCS-1 to MCS-4) is supported.
Add header structs to rlc.h and extend
Encoding::rlc_write_dl_data_header accordingly.
Sponsored-by: On-Waves ehf
|
|
Currently the checks for and the actual polling is done in several
places by copy & paste of several lines of code. This hinders changes
of they polling is handled internally and also is likely source of
programming mistakes.
Separate this into a check_polling function, that checks whether
polling is possible and returns the FN and the RRBP to be used in
that case. Otherwise the cause is logged (LOGL_DEBUG) and a negative
error value is returned. There are no other side effect beside the
logging.
If the call is successful, the set_polling method can be used to
actually register the polling.
Extend the encoder functions' parameters lists by an rrbp parameter.
Sponsored-by: On-Waves ehf
|
|
Currently tbf->control_ts is used to look up an incoming poll
response. Since that may eventually change, poll timeouts could
theoretically happen in that case.
Store the real poll TS in tbf->poll_ts at all places where
tbf->poll_fn is set. Do not use tbf->control_ts to look up
outstanding polls.
Sponsored-by: On-Waves ehf
|
|
Currently there are some places where tbf->control_ts != ts is
evaluated to check, whether ts is a control slot.
Replace these expressions by tbf->is_control_ts(ts) which does
the same whitout exposing internal fields.
Sponsored-by: On-Waves ehf
|
|
Currently the is a release() function which takes care of
statistics and shutdown of properly finished TBFs, but which
cannot be used for aborted TBFs.
This commit add an abort() method which handles unacked RLC blocks
as lost and doesn't start a release timer. This method will be
invoked by tbf_free() for downlink TBF.
Sponsored-by: On-Waves ehf
|
|
Currently the values in the BSSGP RA Cap IE are eventually use
overwrite the 'good' values obtained from the MS earlier.
Use the 'good' values when the are present, which is assumed if
at least one of ms->ms_class() or ms->egprs_ms_class() is set.
Sponsored-by: On-Waves ehf
|
|
The bitvec RBB is always valid, even when FINAL_ACK_INDICATION is
set. This is done by the ack/nack decoder function which fake a
bitmap starting with V(A) up to V(S)-1 in that case (see
Decoding::handle_final_ack).
Call gprs_rlcmac_dl_tbf::update_window unconditionally and only use
is_final for logging and TBF state changes in
gprs_rlcmac_dl_tbf::rcvd_dl_ack.
Sponsored-by: On-Waves ehf
|
|
Currently a window size of 64 is being hard coded.
Use the length of the show_rbb string instead.
Sponsored-by: On-Waves ehf
|
|
The current methods are based on the GRPS specific RBB and SSN values
and formats which are not compatible with EGPRS.
Add a second set of similar methods with the same semantics but
which are based on a bitvec and the first BSN instead.
The following methods are affected:
- gprs_rlc_dl_window::update
- gprs_rlcmac_dl_tbf::rcvd_dl_ack
- gprs_rlcmac_dl_tbf::update_window
Sponsored-by: On-Waves ehf
|
|
The MS' RLC receiver needs a valid CPS field to decode the block.
Add the function gprs_rlc_mcs_cps to generate the CPS value based on
coding scheme, puncturing, and padding.
Sponsored-by: On-Waves ehf
|
|
Currently the GPRS data block encoding is applied to every
coding scheme, even if an MCS is selected.
This commit renames the actual encoding function to
rlc_data_to_dl_append_gprs (not exported) and puts
selection code into Encoding::rlc_data_to_dl_append. This
requires an additional cs argument.
Sponsored-by: On-Waves ehf
|
|
Currently TBF related tasks (status changes, counter updates,
LLC dummy command insertion) as well as
RLC data block generation are done within create_new_bsn.
The data block creation part has already been copied to
the stateless Encoding::rlc_data_to_dl_append function.
This commit changes create_new_bsn to use the encoder function and
just care about the TBF related stuff.
Since the rlc_data_to_dl_append function has been validated against
the test cases being described in annex B of TS 44.060, this
commit fixes an encoder bug which leads to broken LLC frames in
a special case (see example B.2, the main header's E bit was
erroneously set to 1 in that case). When this happens, the LLC frame
will get discarded after reassembly, so that TCP will have to
retransmit the lost packet.
Sponsored-by: On-Waves ehf
|
|
This commit removes the use of struct rlc_dl_header from
gprs_rlcmac_dl_tbf::create_dl_acked_block and
gprs_rlcmac_dl_tbf::create_new_bsn. Instead of patching the
data area directly, the RLC block encoding functions are used.
Note that the data unit encoding is still hard-coded to GPRS in
create_new_bsn, so using MCS 1-4 (albeit being supported by the
encoder) will not work yet.
Sponsored-by: On-Waves ehf
|
|
Currently the RLC message length that is obtained from the DSP is
reduced by 1 if the last byte of the buffer includes spare bits.
While this worked well with GPRS, these bits are being used to
encode RLC blocks in EGPRS mode. Thus this last byte must not be
chopped off. The functionality of the code is not affected by this,
since the modified length value is not used.
This commit adds GprsCodingScheme::usedSizeDL/UL to return
the number of bytes needed to encode the message block. If there are
single bits at the end that are to be used (EGPRS), the functions
return the number of full bytes plus 1 (which is the buffer size
reported by the DSP and returned by sizeUL/sizeDL). The commit also
removes the len parameter from rcv_data_block_acknowledged.
Sponsored-by: On-Waves ehf
|
|
The EGPRS multi-slot class need to be parsed from the CSN.1 RA
capability (see gprs_bssgp_pcu_rx_dl_ud).
This commit adds a workaround to get the EGPRS MS class from the MS
object if that is present.
Sponsored-by: On-Waves ehf
|
|
Currently the TBF and MS object use a plain integer value
(current_cs) to manage the coding scheme. This makes it difficult to
support the MCS schemes. GprsCodingScheme supports a partial ordering
of these values (CS and MCS) and provides safe increment and
decrement methods.
Use the GprsCodingScheme type instead of integer for cs fields and
variables. Add a 'mode' to GprsMs which can be set to either GPRS,
EGPRS, or EGPRS_GMSK which also set the initial values of
current_cs_ul/dl. Select the mode based on max_mcs_ul and max_mcs_dl.
Sponsored-by: On-Waves ehf
|
|
Add a few new operators and methods to support the use of
GprsCodingScheme instead of the plain integer currently used.
Sponsored-by: On-Waves ehf
|
|
Currently there is only a mod_sns() method which is being used in
expression like bsn_expr & win.mod_sns(). This only works, because
it is known that mod_sns() is (sns() - 1) where sns() in turn is
always 2^n. This is error prone, hard to read, and relies on window
specifics that should be kept inside the respective module.
This commit adds a mod_sns(uint bsn) method to gprs_rlc_ul_window and
gprs_rlc_dl_window, that encapsulates and hides this optimized
computation.
Sponsored-by: On-Waves ehf
|
|
Currently the coding scene is stored as number N, where there scheme
is CS-N.
This commit replaces this by a GprsCodingScheme type cs value. The
gprs_rlcmac_cs table is no longer needed and thus removed.
Sponsored-by: On-Waves ehf
|
|
The multislot (MS) class and the EGPRS MS class can also be passed
via BSSGP in an MS Radio Access Capability element which can
optionally be contained in a DL-UNITDATA PDU. While this case is fully
supported for GPRS, the EGPRS MS class in BSSGP messages is ignored.
This commit extends gprs_rlcmac_dl_tbf::handle to pass the EGPRS MS
class, too.
Note, that the EGPRS class is not yet taken from the CSN.1 RA
capability and is always set to 0. Note also, that append_data
still uses ms_class only.
Sponsored-by: On-Waves ehf
|
|
Add an egprs_ms_class argument to the allocation functions and
set/pass it where necessary.
Sponsored-by: On-Waves ehf
|
|
The leak rate sent to the SGSN does not reflect the current CS level,
lost frames, and control message overhead. So the SGSN cannot do
proper queue control under non-optimal conditions.
This commit computes the leak rate for the last flow control interval
by computing the maximum theoretical leak rate and basically
substracting control blocks, nacked blocks, and reduced block sizes
due to CS downgrade. By using this approach, the value will by more
stable on low load, where the value will tend to be near the value
derived from the configuration. On full load the transmitted value is
completely derived from the measurements.
Note that the MS default values are no adapted to the adapted BVC
leak rate, since a single MS which has a lower link quality would
otherwise be reducing the rate of another MS with good radio
conditions, which would not make much sense if they did not share any
PDCH.
Sponsored-by: On-Waves ehf
|
|
Currently reuse_tbf (partly) resets the old DL TBF and uses its PACCH
to establish a new DL TBF. The method can not be used with UL TBFs.
This commit replaces the reuse_tbf method into a
gprs_rlcmac_dl_tbf:release method which triggers the TBF's timer
based deletion (so that the TFI is still reserved for some time) and
a gprs_rlcmac_tbf::establish_dl_tbf_on_pacch which can establish DL
TBFs on existing PACCHs of either DL or UL TBFs.
Sponsored-by: On-Waves ehf
|
|
This commit adds the relevant frame number to the "poll timeout"
logging message. In addition, logging is added to the places where
poll_fn gets set.
The goal is to track down the source for frequent "poll timeout"
messages.
Sponsored-by: On-Waves ehf
|
|
While this does not happen in real use, and unset btcx can lead to
segfaults in test cases. The other code outside of gprs_bssgp_pcu.cpp
does not depend on bctx being non-NULL:
Sponsored-by: On-Waves ehf
|
|
Currently the merging of the meta information (MS class, IMSI) takes
place in gprs_rlcmac_tbf::merge_and_clear_ms(). This makes it
difficult to merge the internal state and does not directly relate to
TBFs anyway.
This commit moves this into a new method GprsMs::merge_old_ms.
Sponsored-by: On-Waves ehf
|
|
The confirm_tlli method does not handle TLLI clashes in the MS
storage.
This commit changes gprs_rlcmac_dl_tbf::handle() to use update_ms
instead.
Sponsored-by: On-Waves ehf
|
|
When doing an RA Update the network can request to change the TLLI.
In this case, there can be 2 MS objects with different TLLI for a
single real MS. The first is associated with the old TLLI and the
IMSI, while the second is associated with the new TLLI and no IMSI if
it had been created for the uplink TBF. When the first message with
the new TLLI and the IMSI arrives from the network, the PCU is able
to detect this.
Currently this is not handled properly. The TBFs of the old MS object
are not cleaned up properly, keeping the old MS from being deleted.
This patch modifies gprs_rlcmac_dl_tbf::handle to check for this and
if neccessary to move an existing DL TBF and to clean up the old MS
object to ensure its deletion.
Sponsored-by: On-Waves ehf
|
|
Currently when receiving a PACKET DL ACK/NACK message with the Final
Ack Indicator bit set, the TBF's state is set to
GPRS_RLCMAC_WAIT_RELEASE but T3193 is only started when the LLC queue is
empty. Otherwise the reuse_tbf() method is called to establish a new
DL TBF. In that case, the timer is not started. This will leave the
current TBF without a timer so it is potentially not released later
on.
This is recognisable by sticky entries in the output of the
'show tbf all' command and possibly allocation failures if there are
too many of them.
This commit changes the code to always start T3193 to make sure, that
a timer is always active when the the state is set to
GPRS_RLCMAC_WAIT_RELEASE.
Note that TS 44.060, 9.3.2.6 requests to release the 'old' TBF
immediately in some cases, which is not implemented by this change.
This will lead to a longer reservation period of the TFI only, which
is safer than reassigning it too early.
Sponsored-by: On-Waves ehf
|
|
Currently packets are only dropped if they have reached their maximum
life time. This leads to LLC queues being constantly filled under
load, increasing the latency up to the maximum life time. This kind
of bufferbloat hinders TCP's congestion avoidance algorithms. To keep
the queues short, the CoDel active queue management algorithm can be
used.
This commit changes to llc_dequeue method to apply the CoDel
algorithm to selectively drop LLC frames before they passed to the
TBF layer to be encoded in BSNs. This feature is currently disabled
by default.
The CoDel state is managed per MS since the LLC queues are also kept
in the MS objects.
Note that there is still some buffering in the TBF objects, in the
worst case (CS4) 3.5kByte + LLC-MTU octets are stored there. The
resulting additional packet delay is not (yet) taken into account for
CoDel.
Also note that configuration changes are applied to new MS objects
only.
The following VTY commands are added to the 'pcu' node:
- queue codel activates CoDel, the interval is selected by
the implementation
- queue codel interval <1-1000>
activates CoDel with a fixed interval given
in centiseconds (10ms-10s)
- no queue codel deactivates CoDel
Which interval value to use is still an open issue. For high speed
links (e.g. Ethernet), CoDel suggests 100ms. For slower links, the
expected RTT is recommended. The current implementation uses a
default value of 2000ms.
Measurements:
Note that the following measurements depend on several other factors,
most notably the interaction with the SGSN's flow control. They are
just examples to give an idea how CoDel might influence some
parameters.
The measurements have been done with a single E71, first with a
running ping only (Idle), then with an additional TCP download
of a 360k file (Busy). The CoDel interval was set to 1s.
- Idle :
ping ~400ms, avg queue delay 0ms, dropped 0
- Busy, No CoDel:
ping ~6s, avg queue delay 4-6s,
dropped 0, scheduled 948, duration 54s
- Busy, CoDel:
ping 500-1500ms, avg queue delay ~600ms,
dropped 77, scheduled 1040, duration 60s
More measurements with two MS downloading in parallel (two
independant measurements per case).
- Busy, No CoDel:
dropped 0, scheduled 1883, duration 121s
dropped 19, scheduled 2003, duration 133s
- Busy, CoDel:
dropped 22, scheduled 1926, duration 116s
dropped 22, scheduled 1955, duration 108s
Sponsored-by: On-Waves ehf
|
|
Currently the TFI and the TRX have to be determined before the actual TBF
allocation function is called, passing TFI and TRX number as
parameters. This does fit to TFI reuse for different slots, since
this were tightly coupled with the slot selection.
This commit just moves the TFI selection into the alloc_algorithm
functions. The tfi parameter is removed from the the TFI alloc
functions. The trx parameter is changed into use_trx to optionally
limit the trx selection (same semantics like in tfi_find_free).
Sponsored-by: On-Waves ehf
|
|
Since both TBF are based on the same reservation which means that
they should be compatible with respect to the slot usage, and since
the new TBF has not been forced to single slot usage, an update of
the allocation is not necessary now.
This commit removes the call to update() from within reuse_tbf().
Sponsored-by: On-Waves ehf
|
|
The call to tbf_alloc_dl_tbf misses the pointer to the GprsMs object
which is already known in that case (tbf_reuse). This leads to a full
reallocation of the PDCH slots, which is possibly incompatible with
the old set of slots. This can result in hanging TCP connections and
TCP connection failures.
This commit replaces the old NULL value by the actual GprsMs object.
Since the set_ms() is also done within the tbf_alloc_dl_tbf method,
that call is removed.
Sponsored-by: On-Waves ehf
|
|
Since set_ms() is caled on the new DL TBF, the old DL TBF loses the
reference to the MS object. This will lead to a segfault, when
update() is called in reuse_tbf().
This commit adds an optional GprsMs* parameter to update() and uses it
for the slot allocation.
This fixes a TbfTest crash that would otherwise occur after applying
the next commit.
Sponsored-by: On-Waves ehf
|
|
Currently the old TBF (either uplink or downlink) is passed around at
TBF allocation mainly to get information about the MS. To implement
more complex allocation algorithms, the MS object itself will be
needed anyway.
This commit replaces the old_tbf arguments by MS object arguments.
Sponsored-by: On-Waves ehf
|
|
Currently the receive and expiry timestamps are prepended to the LLC
msgb before it is passed to gprs_llc_queue::enqueue(). Since this meta
information should not be counted as LLC octets, the gprs_llc_queue
needs to known about this (unless the correction was done in the LLC
layer).
This commit moves the meta information storage code into
gprs_llc_queue. The meta data is now stored in the control block
(cb) area of the msgb.
Note that the info pointer that is returned from the dequeue method
is only valid if that method returns a (non-NULL) msgb. It must not
be used after that msgb has been modified or freed.
Sponsored-by: On-Waves ehf
|
|
When the MS is pinged with a longer interval, many packets get lost
even if the GprsMs object is kept. If the interval is above the time
where the DL TBF is in state FLOW (mainly influenced be the
dl-tbf-idle-time command), an new TBF must be requested via AGCH for
each ICMP PING message.
Currently the LLC frame containing the PING is immediately stored
in the TBF and gets lost, if TBF establishment fails for some reason.
This commit moves all calls to put_frame() to schedule_next_frame(),
where the data is moved from the LLC queue to the frame storage
within the TBF object. This method is only called from within
create_new_bsn() when the TBF is in the FLOW state and the frame is
going to be encoded immediately.
At all other places, where put_frame() has been called before, the
LLC message is just appended to the LLC queue in the GprsMs object.
This change effectively simplifies the related code parts, since
date/len information and discard notifications is no longer needed
there.
Ticket: #1759
Sponsored-by: On-Waves ehf
|
|
This change lets the test suite fail, so it get its own commit.
Sponsored-by: On-Waves ehf
|
|
Currently the CS level gets changed quickly if single RLC/MAC blocks
are sent (e.g. LLC dummy commands), since the rate is either 0% or
100%.
This commit ignores measurements which are based on a single block
only.
Sponsored-by: On-Waves ehf
|
|
To help with the debugging, optimisation, and understanding of this
method, this commit adds an info string containing a flag character per
RLC/MAC data block in the current window.
Note that the blocks are shown in reversed order (highest BSN first)
in comparison to other logging output.
Sponsored-by: On-Waves ehf
|
|
To cope with transmission failures due to bad radio conditions, a
different coding scheme with more redundance can be used.
This commit adds an implemenation that is based on the Ack/Nack
ratio per PACKET DOWNLINK ACK/NACK message received from the MS.
Basically the CS level is decreased, if the block error rate goes
above cs_adj_upper_limit (default 33%), and it is increased, if the
rate drops below cs_adj_lower_limit (default 10%). Only blocks that
have been encoded with the current CS are taken into account.
Note that this approach doesn't measure the MS->BTS conditions and
that the measurement values reported by the MS are not taken into
account.
Ticket: #1739
Sponsored-by: On-Waves ehf
|
|
Currently the current CS value is stored in the cs field of
gprs_rlcmac_tbf and initialised when it is used the first time.
This commit adds separate fields for UL and DL CS values to the
GprsMs class and provides corresponding getter methods for GprsMs and
gprs_rlcmac_tbf.
Ticket: #1739
Sponsored-by: On-Waves ehf
|
|
The ms_class value is a property of the MS and thus belongs to the
GprsMs class. Nevertheless the MS object is created after the TLLI
gets known, so the value still has to be stored in the TBF initially.
This commit add the ms_class value to the GprsMs class and introduces
TBF accessor functions which either access that object or, if that is
not available, the value stored locally.
Ticket: #1674
Sponsored-by: On-Waves ehf
|
|
Currently the enqueued DL LLC messages which have not yet passed to
RLC/MAC encoding are eventually copied from one TBF to the next one
(see gprs_rlcmac_dl_tbf::reuse_tbf). Since the enqueued LLC messages
are related to a specific MS, they should be stored at that layer.
This commit moves the gprs_llc_queue object to GprsMs and changes the
TBF's accessor methods accordingly. The LLC copying code is removed
from gprs_rlcmac_dl_tbf::reuse_tbf().
Sponsored-by: On-Waves ehf
|
