This fixes TBF objects leaking and ending up alive when the MS object is
explicitly freed through talloc_free (and sporadically
crashing TbfTest once a timeout for them occur).
This mostly affects unit tests, where most of the explicit free()
happens.
In osmo-pcu, in general, the GprsMs object only gets _free() called when
its resource count reaches 0, aka no more TBFs are attached to it. Hence
in general GprsMs object is freed() only when no TBFs (to be leaked) are
present.
However, in the unit tests it's usual that we want to wipe the entire
context by eg. feeing the PCU, the BTS or MS object, which should also
free the related TBFs.
When running osmo-pcu this may only be an issue when the MS object is
freed explicitly, which could happen for instance when a BTS is torn down,
ie. PCUIF going down, moment at which all GprsMs of that BTS are freed.
But in there actually it iterates over PDCHs to free all TBFs, so it's
fine.
If we iterated over MS, this could have ended up in a crash, like
it happened in TbfTest sporadically, but it's not a bit problem if we
crash + restart at that time since anyway the BTS is gone ore just
getting up around that time.
Related: OS#6359
Change-Id: Ibbdec94acb8132be20508d3178d88da44bfaf91d
Change format to print the state at the end, to resemble more the same
format used by FSMs.
Furthermore, by moving it at the end, print it only when "enclousure" is
requested, aka when not requested by FSM to update its internal name.
The consequence of this logc is that log lines printed from FSM don't
end up with the same state string printed twice in different places.
While at it, shorten the EGPRS/GPRS indicator to one character, which
should be understandable enough since it matches what's usually seen in
mobile phones to signal one or another.
Change-Id: I86b5f042fae77721b22fc026228677bd56768ba9
This patch finally decouples TBF allocation from resource allocation.
This will allow in the future reserving resources without having to
require a TBF object to exist.
Change-Id: I2856c946cb62d6e5372a1099b60e5f3456eb8fd4
This commit changes lots of stuff in the MS release lifecycle, but
there's no really good way to split this into patches which make sense,
since all the chaos is intensively entangled.
Get rid of the ms_callback complex mess, it is not needed at all.
Previous MS release was strange due to the existance of previous
ms_callback.idle concept and MS storage: the MS signalled when it went
idle (no TBFs attached) and waited for somebody outside to free it,
while then arming itself the release timer to release itself if it was
not released by whoever.
The new lifecycle follows an easier (expected) approach: Whenever all
TBFs become detached from the MS and it becomes idle (use_count becomes
0), then it frees its reserved resources (TFI, etc.) and either:
* frees itself immediatelly under certain conditions (release timeout
configured = 0 or MS garbage with TLLI=GSM_RESERVED_TMSI)
* Arms release_timer and frees itself when it triggers.
If during release_timer the MS is required again (for instance because a
new TBF with TLLI/IMSI of the MS is observed), then a TBF is attached to
the MS and it is considered to become active again, hence the release_timer
is stopped.
OS#6002
Change-Id: Ibe5115bc15bb4d76026918adc1be79469c2f4839
This allows having full information on the control TS easily reachable
(like TRX ofthe PDCH), and makes it easy to compare TS by simply
matching the pointer address.
Change-Id: I6a97b6528b2f9d78dfbca8fb97ab7c621f777fc7
The triggering of the assignment for the new TBF allocation is part of
the "upgrade to multislot" process, hence move that inside the function.
Change-Id: Ic2b959f2476b900cb263ccd8f6698ed843bafd29
The 2 types of TBF share some parts of the implementation, but actually
half of the code is specific to one direction or another.
Since FSM are becoming (and will become even) more complex, split the
FSM implementation into 2 separate FSMs, one for each direction.
The FSM interface is kept compatible (events and states), hence code can
still operate on the tbfs on events and states which are shared.
Test output changes are mainly due to:
* FSM instance now created in subclass constructor, so order of log
lines during constructor of parent class and subclass slightly
changes.
* osmo_fsm doesn't allow having 2 FSM types with same name registered,
hence the "TBF" name has to be changed to "DL_TBF" and "UL_TBF" for
each FSM class.
* FSM classes now use DTBFUL and DTBFDL instead of DTBF for logging. Some
tests don't have those categories enabled, hence some log lines
dissappear (it's actually fine we don't care about those in the test
cases).
Change-Id: I879af3f4b3e330687d498cd59f5f4933d6ca56f0
Use same formatting similar to what's now used in TBF, which is far more
easy to grep and follow. This way one can easily follow what happens to
a given IMSI, a give TFI, a given TLLI, etc.
Change-Id: If9b325764c8fd540d60b6419f32223fd7f5a5898
use a format in tbf::name() which is sanitized (osmo_sanitize) and hence
can be used both in regular log as well as for its internal FSM ids.
Until now, the FSMs contained a small amount of information with
different formatting than the regular LOGPTFB(), which made it difficult
to grep or follow a TBF through its lifetime looking at logs. The new
unified format makes that a lot easier.
Extra information is now printed if available, such as IMSI.
Furthermore, the TFI is updated to include BTS and TRX, since the TFI is
unique within the scope of a TRX.
Change-Id: I3ce1f53942a2f881d0adadd6e5643f5cdf6e31da
This commit actually addresses 2 errors:
1- gprs_bssgp_pcu_rx_paging_ps() called gprs_rlcmac_paging_request()
with MI which can be either TMSI or IMSI, and the later always called
bts_pch_timer_start() passing mi->imsi regardless of the MI type. Hence,
trash was being accessed & stored into bts_pch_timer structures if MI
type used for paging was TMSI.
2- When the MS received the PS paging on CCCH and requests an UL TBF, it
will send some data. If one phase access is used for whatever reason,
the IMSI may not be yet available in the GprsMs object since we never
received it (and we'd only have it by means of PktResourceReq). Hence,
let's better first try to match the paging by TLLI/TMSI if set in both
places, and otherwise use the IMSI.
Related: OS#5297
Change-Id: Iedffb7c6978a3faf0fc26ce2181dde9791a8b6f4
This allows distinguishing when a TBF didn't set the TFI. Useful to
identify dummy reject TBFs, etc, and make sure a non-dummy TBF set its
TFI properly.
Change-Id: Iecf54a24041bd14f4ef5b86e57c3732e1b69d463
This helps distinguishing the case where a TBF is in the initial state
and the unexpected case where osmo_fsm_inst_state_name reports "NULL"
due to fi pointer being NULL.
Change-Id: Ieaabfc9fa0dedb299bcf4541783cf80e366a88c3
We are freeing the object immediately afterwards anyway, so no need to
pretend it went through the normal state release.
Leaving current state as it is actually provides more information on
what was the status/state at the time the TBF had to be freed.
Change-Id: I3016caaccc2c43e1e300f3c6042d69f8adcd9d69
At some point later in time the state_flags will most probably be split
into different variables, one ending up in a different FSM. It is moved
so far to the exsiting FSM from the C++ class since it's easier to
access it from C and C++ code, and anyway that kind of information
belongs to the FSM.
Related: OS#2709
Change-Id: I3c62e9e83965cb28065338733f182863e54d7474
Run bts_pch_timer_remove() on each entry of the BTS specific pch_timer
list, so we don't have a memory leak and so the timer doesn't
potentially fire for a deallocated BTS.
Fixes: d3c7591 ("Add counters: pcu.bts.N.pch.requests.timeout")
Change-Id: Ia5e33d1894408e93a51c452002ef2f5758808269
This is only an initial implementation, where all state changes are
still done outside the FSM itself.
The idea is to do the move in several commits so that they can be
digested better in logical steps and avoid major break up.
Related: OS#2709
Change-Id: I6bb4baea2dee191ba5bbcbec2ea9dcf681aa1237
Before this patch, allocate_usf() was implemented to only allocate 1 USF
per TBF, regardless of the available ul_slot mask.
As a result, only 1 slot at max was allocated to any TBF. That's a pity
because usual multislot classes like 12 support up to 2 UL slots per TBF
(in common TS with DL).
This patch reworks allocate_usf() to allocate as many UL multislots as
possible (given mslot class, current USF availability, TFI availability,
related DL TBF slots for the same MS, etc.).
As a result, it can be seen that AllocTest results change substantially
and maximum concurrent TBF allocation drops under some conditions.
That happens due to more USFs being reserved (because each TBF has now
more UL slots reserved). Hence now USF exhaustion becomes the usual
limitation factor as per the number of concurrent TBFs than can be
handled per TRX (as opposed to TFIs previously).
Some of the biggest limitations in test appear though because really
high end multislot classes are used, which can consume high volumes of
UL slots (USFs), and which are probably not the most extended devices in
the field.
Moreover, in general the curren timeslot allocator for a given
multislot class will in general try to optimize the DL side gathering
most of the possible timeslots there. That means, for instance on ms
class 12 (4 Tx, 4Rx, 5 Sum), 4 DL slots and 1 UL slot will still be
selected. But in the case where only 3 PDCHs are available, then with
this new multi-slot UL support a TBF will reserve 3 DL slots and 2 UL
slots, while before this patch it would only taken 1 UL slot instead of
2.
This USF exhaustion situation can be improved in the future by
parametrizing (VTY command?) the maximum amount of UL slots that a TBF
can reserve, making for instance a default value of 2, meaning usual
classes can gather up 2 UL timelosts at a time while forbidding high-end
hungry classes to gather up to 8 UL timeslots.
Another approach would be to dynamically limit the amount of allowed
reservable UL timeslots based on current USF reservation load.
Related: OS#2282
Change-Id: Id97cc6e3b769511b591b1694549e0dac55227c43
When allocating multiple slots for a UE the following example
is not allowed 'UU----UU' for a UE class 12.
The time slot number can not roll over 7 and move to 0.
44.060 or 45.002 only specifies contigous however it was unclear
it this is an allowed pattern.
Only the example 45.002 B.3 in release 12 cleared this up.
It gives an example for a multi slot class 5 UE which has 7 possible
configuration this means the rolled over is not allowed.
Multislot class type 2 UE doesn't have this limitation.
Further if a UE supports 8 time slots this is not a limitation because
the window size (45.002 B.1) can include all time slots.
Releated: SYS#5073
Change-Id: I16019bdbe741b37b83b62749b840a3b7f4ddc6c7
The assumption that TLLI 0x00000000 is invalid and can be used
as the initializer is wrong. Similar to TMSI, 0x00000000 is a
perfectly valid value, while 0xffffffff is reserved - use it.
According to 3GPP TS 23.003, section 2.4, a TMSI/P-TMSI with
all 32 bits equal to 1 is special and shall not be allocated by
the network. The reason is that it must be stored on the SIM,
where 'ff'O represents the erased state. According to section
2.6 of the same document, a local/foreign TLLI is derived from
P-TMSI, so the same rule applies to TLLI.
I manually checked and corrected all occurances of 'tlli' in the
code. The test expectations have been adjusted with this command:
$ find tests/ -name "*.err" | xargs sed -i "s/0x00000000/0xffffffff/g"
so there should be no behavior change. The only exception is
the 'TypesTest', where TLLI 0xffffffff is being encoded and
expected in the hexdump, so I regenerated the test output.
Change-Id: Ie89fab75ecc1d8b5e238d3ff214ea7ac830b68b5
Related: OS#4844
Some tests were wrong (TypesTest) and required modification, since they
were setting a EGPRS MS but then expecting a GPRS assignment.
Change-Id: I9d3ee21c765054a36bd22352e48bde5ffca9225a
This is another step forward towards a more clear data model where a TBF
always has a MS object (which may be lacking some information, and at a
later point when more information is found, it may actually be a
duplicated MS object and hence one duplicate removed and the TBF moved
to the object being kept).
This helps for instance in removing duplicated information stored in
the TBF which is really per MS, like ms_class, ta, etc. Since there's
always a MS object there's no need to keep a duplicate in both classes
in case there's no MS object.
It can already be seen looking at unit test logging that this kind of
data model already provides better information.
Some unit test parts were needed to adapt to the new model too.
Change-Id: I3cdf4d53e222777d5a2bf4c5aad3a7414105f14c
In those cases since a string pointer was passed, it always printed
"single" instead of whatever really was being used, since the string
pointer was not NULL.
Change-Id: Idab7d18e8f519e10fc3df4007634661c46f9256d
If TS allocation fails due to unavailable TFI, print TRX which was
suggested to allocator. This simplifies allocator debugging but requires
cosmetic modifications to test output.
Change-Id: Icaf97d71d71985d52dc0bda448c26b19fe5645e7
Related: OS#2282
The table B.1 is copy-pasted from 3GPP TS 45.002 and reformatted via
Emacs macros into C struct to avoid typos. The test output expanded
accordingly.
The allocation test expectations and output are adjusted accordingly.
Note: classes 35-45 which need TA offset are not properly supported
yet. This can be extended once we have such devices available for tests.
Change-Id: I1ef2eb99c517f25e7d1e71b985a3e0eb3879eb2c
Related: OS#2282
So far the allocation was only tested up to hardcoded MS class 29. Drop
that assumption and test for all supported MS classes. Adjust expected
test output as necessary.
Note: using mslot_class_max() forces allocation for MS classes 30 and 31
for which no actual data is available (will be added in follow-up
patches) which current implementation treats differently depending on
TX/RX direction - see gprs_alloc_max_dl_slots_per_ms(). Because of that
we have to adjust the expected number of allocations in
test_successive_allocation() as well.
Change-Id: I7737f303d97197ef159b14a19c3312a11f07b433
Related: OS#2282
Previously all TBF-related events were logged as part of DRLCMAC which
is too broad to make it practically useful due to excessive amount of
log messages generated. Introduce dedicated log categories for
TBF-related events. Adjust test output as necessary.
Change-Id: I64d660e5971263d5c63d2ba95d50625c16a594aa
Currently there is no persistent state being used in
alloc_algorithm_dynamic. So algorithm B is even used in persistent
high usage scenarios. If there are many active TBFs, multislot
assigments are not fair, because MS of a "higher" multislot class get
higher troughputs. On the other hand, as long as all PDCH are busy no
bandwidth will be wasted even if all MS use algorithm A.
This commit modifies alloc_algorithm_dynamic to disable algorithm B
when that call fails. It then keeps it disabled until there is a
single PDCH which is idle (it is considered idle, if there is at most
one active DL TBF assigned to it).
Sponsored-by: On-Waves ehf
The idea behind this meta algorithm is to automatically select one of
the other algorithms based on the system state. Basically algorithm B
will be selected if the PDCH usage is low to improve throughput and
latency. Algorithm A will be selected to support more concurrent MS.
This commit adds a first simple state-less version of this algorithm
that always tries B first and only if that fails A is tried
afterwards.
The following VTY command is added to the 'pcu' node:
- alloc-algorithm dynamic
Ticket: #1934
Sponsored-by: On-Waves ehf
Currently these algorithms modify other objects (MS, TBF, PDCH) even
if the allocation will fail later on. To implement an algorithm that
dynamically tries another algorithm on failure (e.g. A after B), the
first (failing) algorithm should not change or damage anything.
This commit refactors algorithm A and B to delay the actual allocation
until it is known that the allocation will not fail.
Ticket: #1934
Sponsored-by: On-Waves ehf
MS iniated TCP connections generally result in a sequence
of short time UL and longer lasting DL TBFs, being interleaved
between several MS. This scenario is not covered by the existing
tests.
This commit adds a test, that allocates as man as possible TBFs
several times with different test modes without clearing the BTS (and
thus the TBF list) in between. The number of allocated DL TBFs in
each round is expected to be constant.
Sponsored-by: On-Waves ehf
Currently the TFI are managed per TRX, thus only a maximum of 32 TBF
per direction and per TRX are possible simultaneously.
This commit modifies algorithm_a() to allow the sharing of TFI
between different PDCH. Since algorithm A only assigns a single slot
to each TBF, the TFI of each PDCH can be assigned independently.
This increases the maximum to 32 TBF per direction and per PDCH
concerning the TFI allocation.
Ticket: #1793
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
Currently a single bit set is used to maintain a set of used TFI
without distinguishing between uplink and downlink. Since the
namespaces of UL and DL TFI are separate, this implementation is
not correct.
This commit changes gprs_rlcmac_pdch to use a separate bit set for
each direction. It also replace the corresponding conditional fprintf
statement in check_tfi_usage (AllocTest.cpp) by an equivalent
OSMO_ASSERT.
Sponsored-by: On-Waves ehf
This commit adds the check_tfi_usage function that checks the TFI
usage. It iterates through all TBFs, records on which PDCH it uses
which TFI and check for conflicts. It also checks the bits returned
by pdch->assigned_tfi(). The latter suffers from an bug in that
method (no separation of uplink and downlink), so a conditional
fprintf is used instead of an assertion. The method tfi_find_free
is checked for conflicts after allocations.
Sponsored-by: On-Waves ehf
Currently the slot selection of algorithm A is based on the current
slot usage by active TBF. Especially in the Dl after UL case which
reflects the commen use case "MS initiates TCP connection", the
resulting distribution is not optimal with respect to PDCH usage.
This commit changes the implementation to use the slot reservation
information instead.
Sponsored-by: On-Waves ehf
Currently when using the test modes TEST_MODE_DL_AFTER_UL or
TEST_MODE_UL_AFTER_DL, the first TBF is deleted before the second is
allocated. The far more interesting case were to keep the first TBF a
little bit longer until the second TBF has been created and delete
then. This comes closer the the situation observed with real MS,
where the first TBF takes some time (timeout or waiting for Ack)
before it gets deleted and thus detached from the MS object.
This commit delays the call to tbf_free accordingly.
The effect can be observed in the results of the algo A tests, where
the uniform distribution of the allocated PDCH is lost.
Sponsored-by: On-Waves ehf
Currently the search of the "best" slot combination is done
separately from the UL slot selection, which can lead to an
allocation failure due to USF exhaustion even if another combination
had been possible.
This commit reduces the probability for this event by skipping UL
slots without free USF while calculation the capacity.
Note that the implementation is rather inefficient which will be
fixed by the following commits.
Sponsored-by: On-Waves ehf
Since currently the algorithm B will only allocate a single UL slot
and will have to stick to it (first common TS), the other possible UL
slots will not be allocated while the reservation is kept.
This commit adds code to update the reserved set of UL slots to only
reserve the single common TS when the UL TBF is allocated.
Interestingly this leads to fewer allocated TBF in some cases due to
USF exhaustion. This will be improved by the following commit "alloc:
Skip common TS without free USF".
Sponsored-by: On-Waves ehf
Currently al possible UL slots are included in the capacity
calculation which is the base of the slot selection. Nevertheless
UL-only slots will never be used, since only one uplink slot (which
must be a common slot) will be used.
This patch changes the code to only include common slots in the
capacity sum.
Note that this might not be optimal if algorithm B supported
multiple uplink slots.
Sponsored-by: On-Waves ehf
The current implementation always starts the downlink slot allocation
with the first possible slot, depending on which channels are enabled
and which multislot class is offered by the MS. So in configurations
with many (>4) PDCH, some PDCH are not really used.
The new implementation introduced by this commit differs as follows:
- The reservation mechanism provided by GprsMs is used to avoid
incompatibilities is used in the same way like algo A does. This
basically means, that the allocation is done once when the first
TBF is requested and then used until all TBF have been released.
- All combinations of Rx and Tx slots are checked for compatibility
with the multiscot class. Basically the combination with the most
usable PDCH and the least number of reservations is used.
- Only one UL slots is provided.
- Tta and Tra are checked.
Sponsored-by: On-Waves ehf
Currently only the first enabled PDCH will be used. Beside the
throughput this will also limit the number of TBFs:
- number of UL TBFs <= 7
- number of DL TBFs <= 32
This commit changes the allocation algorithm to use the PDCH with the
least number of attached TBFs. This will improve the troughput in
both directions and the UL limits:
- number of UL TBFs <= min(32, N_PDCH * 7) UL TBFs
Ticket: #1794
Sponsored-by: On-Waves ehf
This commit adds:
- an assertion to check that count is valid,
- an assertion the verify that tbf_alloc did not fail
- a slots parameter to specify the enabled slots
- additional test invocations with more slots enabled
Sponsored-by: On-Waves ehf
This test allocates as many as possible UL/DL TBF pairs, shows their
PDCH usage, and checks how many of them has been created
successfully.
Sponsored-by: On-Waves ehf
Pau Espin
Oliver Smith
Pau Espin Pedrol
Vadim Yanitskiy