mirror of
https://github.com/MobileGL-Dev/MobileGL
synced 2026-11-17 15:03:25 +09:00
b4f7638742e7875e936246fc44cfc80c41d920dc
44
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
b4f7638742 |
[Perf] (P12): the create window lands as the default - declared-answer retention replaces the ring, and the load frame goes 28.66 s -> 16.52 s (11.02 s at window 4)
The three things that made the deferral inert are all fixed, in the order they were
found: the reply ring could not hold an answer across the ~1,000 records between two
creates; a residency budget could not help because what displaces an answer is what
the SERVER sends, not what the client emits; and so the mechanism had to be retention.
WHAT LANDED
* ILink::DeclareReplyRead + LinkCapabilities::RetainsReplies. The wire declares a
seq will be read BEFORE it publishes, so the declaration cannot race its own
answer; a link with the capability false ignores it, and the window does not
defer on one (CreateWindowEffective() is gated on it).
* StreamLink's ring of the last 8 answers becomes a declared-answer store: a
monotone whitelist plus a cursor, answers kept until their reader TAKES them
(take-and-delete under one lock, bytes swapped into the scratch), and a capacity
of 24 that is FATAL to exceed rather than evicting. An evicted wanted answer is
unrecoverable - re-latching by sending another create wipes the server's record
(PipeApply.cpp:2033) and its PendingUploads (PipeApply.h:332) while the client
cleared its dirty flags at emission (PipeApply.h:298-304) - so the capacity is
sized so it cannot be reached and reaching it says its own name.
* An answer nobody declared is DROPPED, which is what keeps the store bounded by the
client's deferral depth instead of by the server's answer rate. Retain-on-arrival
grows ~11.3 entries per create against a server that ignores the no-reply bit.
* A read that fails after WaitForApplied has returned Reached is now Fatal{ReplyMissing}
naming the seq: on the stream arm the reply frame strictly precedes the progress
frame that publishes appliedSeq, so that combination is a broken invariant, and
retrying it (what the row used to do) wedges the window for the rest of the
process while leaving every deferred object latched on a provisional accept.
* The residency budget is deleted (WireTables.cpp keeps its history), the window
retirement branch drains once before falling back - it could otherwise strand the
answers it was holding - and Detach clears the store.
MEASURED ON THE DEVICE, load frame, both ends rebuilt, frame-sent identical per op
(create 4,536, records 103,097) throughout:
window 1 (R-5's shape) 4,536 takes 28.66 s
window 2 explicit 2,500 17.27 s
window 2 AS THE SHIPPED DEFAULT 2,563 16.52 s
window 4 1,644 11.02 s
window 2 against an OLD SERVER 2,663 16.38 s (zero Fatals)
The last row is the version-skew case the design review named as a must: a server
built before the no-reply bit answers ~55,903 records a frame, the whitelist drops
every answer nobody declared, and the deferral still reads its own back.
GATES (172 + 15 green), and the important one is proved to go red:
* ADeclaredAnswerSurvivesFarMoreInterveningAnswersThanAnyFixedBuffer - reverting the
store to "keep the last 8 arrivals" fails it immediately with "the declared answer
did not survive 64 intervening ones". That is the line between the ring-1024
stand-in and a bounded mechanism.
* AnAnswerNobodyDeclaredIsNotKeptAtAll - red if everything is retained.
* TheWindowRefusesToDeferOnALinkThatCannotRetainAnswers - on the in-process (ShmLink)
arm the knob at 4 must still block every create, red if the capability gate goes.
* The refused-create case now reads the repair that exists on THAT arm (the latch),
and asserts the window's own books are empty there.
NOT DONE, and named: the window has no unit gate on the STREAM arm yet - only the
device runs. Three rounds of this feature were green in the lanes while inert on the
device because the lanes and the device run different transports; the transport-level
retention gates close that for the store, and the window-level one should be built on
StreamClientPublicationTest's real stream session next.
|
||
|
|
0d0a8f9226 |
[Feat] (MG_Remote): a record can tell the server its answer will never be read - the bit works and cuts a frame's replies by 92%, and it proves the window's real blocker is the FIFO drain, not the pool depth
The create window could not collect its deferred answers. The previous commit
diagnosed that as the reply pool being 8 slots deep against 55,903 replies in one
frame, and left a no-reply bit as the way out. This lands that bit, measures it on
the device, and reports that the diagnosis was half wrong.
WHAT LANDED. ClientSession::EmitAndWaitTails takes a second question beside
wantReply: willReadReply. They are not the same question and the window is the case
that separates them - it emits with wantReply=false and READS the answer later, so
its records must still be answered. kRecNoReply (RingRecordFlags bit 5, which
aliases kOptional in the call-flag space; the flag-space asserts were extended and
the alias named) is stamped when willReadReply is false, and the decoder skips
PostReply while keeping every acceptance counter - "nobody reads the answer" is not
"nothing happened". The two illegal combinations are Fatal by name.
Three call sites say it: SetTextureParams for a confirmed object, the respecify
row's fire-and-forget half, and sub-data where MGPipeSubDataWantsItsReply decides.
MEASURED ON THE DEVICE (aikjs4s8cys8ba7l, rebuilt APK + client, both halves new):
server: skipped=50,001 answered=4,521 (a frame's replies: 55,903 -> 4,543)
client: Reply frames received seen=4,501, latest seq=97,181
frame-sent: identical per op, records untouched
create row: by_op 2 = 9,070 (was 4,536) - because the drain's blocking take is
now counted, so the row's number stays honest instead of reading as
"create stopped waiting" while the waiting merely moved
AND THE WINDOW STILL READ BACK NOTHING: readfail 8,951 -> 8,872, i.e. -0.9%.
THE REAL CAUSE, from the miss probe:
P12miss n=1 want=310 replySeq=1504 ring=[1175..1504]
P12miss n=9001 want=310 replySeq=102915 ring=[102866..102915]
want=310 every time. The drain is STRICTLY FIFO and a failed read has no way out -
it retries the same head for ever. Any reply buffer is bounded by design (8 pool
slots, or the 8-frame ring the stream link now keeps), so with 4,536 deferred
answers at least one is overwritten before it is read, and that one wedges the
window permanently. Cutting the churn by two orders of magnitude does not help: it
turns a certainty into a certainty. That is the constraint, not the depth.
THE FIX IS NAMED AND NOT TAKEN (the goal this work ran under stops here on purpose).
On a failed read of the head, abandon it: MGPipeNoteHandleUnpublished +
NoteCreateSuspect + pop - the DECLINE path reused for "unknown". The window then
self-heals through the blocking path and the object merely gets slower. It needs its
own gate: a failed read must let the window ADVANCE.
Gates, all green (172 + 13 + 22 + 103):
PipeWireCodecTest.TheNoReplyBitSkipsTheAnswerAndKeepsTheAcceptanceRecord - the skip,
the kept tally, the answered/skipped split, and the negative control that a record
WITHOUT the bit is still answered (a server that stopped answering everything
would otherwise pass)
RemoteClientControls.ACreateWindowDefersTheAnswersAndTheWindowOffDoesNot - now also
asserts the answers CAME BACK (answers_taken=11, outstanding=1, fallbacks=0 against
12 records). RED-ONCE PROVED: giving the window's creates the literal
"fire-and-forget => skip" predicate makes this case fail.
RingTest - the sixth and last flag alias, and the hand-written exhaustiveness list
extended (with a note that it is only as exhaustive as that list).
Server side: AnsweredRecords()/SkippedAnswers() - the number that made this
debuggable, since a wire change whose effect is invisible is undebuggable.
Client-first is safe: an old server ignores bit 5 and replies as before, so the two
cross-version directions degrade to today's behaviour - this is "both ends must
upgrade to take EFFECT", not "to avoid breaking". Hit that in practice first.
|
||
|
|
a2c1f5f292 |
[Fix] (MG_Remote/Transport): StreamLink::ReadReply answers by seq instead of "is it the newest" - the contract ShmLink already kept, and the reason a deferred answer could never be read back
ILink::ReadReply takes a SEQ (ILink.h:285) and ShmLink implements it against a
ReplySlotPool, so any answer the pool still holds is readable. StreamLink kept ONE
reply - replySeq/replyStatus/reply - and then asked the wrong question of it:
x.replyReady.wait(lock, [&]{ return closed || x.replySeq >= seq || appliedSeq >= seq; });
if (x.replySeq != seq) return ... MOBILEGL_ERR_PROTOCOL_MISMATCH;
The wait is written as "at least this seq"; the check is "exactly this seq". Every
answer older than the newest was therefore unreadable, which nothing noticed because
every caller read the answer it had just waited for - i.e. always the newest one. It
is the transport half of the P12 create window's failure: the window defers up to
four answers and collects them in order, and this link could not hand one back.
THE RING, AT THE POOL'S OWN DEPTH. kReplyRing = kDefaultReplySlotCount (8), because
that is the pool this link stands in for: an answer a reader has not taken by the time
eight newer ones have arrived is one the shm arm has already overwritten too, so the
two transports now lose the same answers and no more. Vectors are sized per reply, so
an idle link pays eight empty ones.
This does NOT make the create window work - on the device it changed nothing, because
the pool itself is 8 slots against 59,673 replies in one frame (see the next commit).
It is kept because the two transports disagreeing about ILink's contract is a bug in
its own right, and because the single slot silently failed any read that raced a newer
reply.
Gate: StreamLinkTest.AnOlderReplyIsStillReadableByItsSeq posts three replies and reads
them oldest-first - the order the window's drain takes them in.
|
||
|
|
1d363eb109 |
[Perf] (MG_Remote): the client counts RECORDS SENT per op beside the waits each row PAID, and the pair says item B stopped the waiting without stopping the sending
- A wait count alone cannot say whether a frame is slow because it sends too much or because it waits too often. The device run made that the next question: one atlas frame paid 43,232 waits while the upload row paid none. - LinkMetricsNoteRecord is called at the encoder's one commit point (PipeWireCodec.cpp, beside the existing CallClass::WireRecords count, for the comment's reason: after chunking no emitter knows how many records its call produced). LinkMetricsRecordsFor reads it back, and every frame that emitted something now logs `P65LinkMetrics kind=frame-sent frame=N records=M by_op=...`. - The host measurement now prints and asserts both halves for the same workload. An atlas-shaped load (64 sub-uploads into one level) sends 128 ResourceSubData records and waits for ZERO of them, while ResourceCreate 76/76, ResourceRespecify 130/130 and SetTextureParams 130/130 are one wait per record. That is item B stated exactly: the records still go out; the round trips do not. It also says what the next thing to fix is - the record COUNT of those three rows, not their waits, which the contract rules they must keep. |
||
|
|
170b789e7f |
[Perf] (MG_Remote/Transport): the per-op reply-wait split prints with every frame, because a whole-session total cannot name the rows behind one frame
- The device run that motivated this made the point: 43,232 of a session's 43,629 reply waits were in ONE atlas frame, and the exit summary never appeared at all - a SIGTERM does not take the session's clean stop path, so LinkMetricsEnd is never reached. The split added one commit ago was therefore invisible on exactly the run that needed it. - Emit now writes a second line for any frame that bought a wait: `P65LinkMetrics kind=frame-op frame=N wait_replies=M by_op=<id>=<count>,...`. The formatting is factored into FormatOpSplit and shared with the exit line. Frames that bought nothing stay quiet, so a steady-state session does not double its log. - Measured on the device (MC 26.2 into a world over TCP, MOBILEGL_PIPE_STATS=1): the steady state is one wait per frame and it is op 9 (Present); the upload-heavy startup frames are `2=50,3=68,47=62` - ResourceCreate / ResourceRespecify / SetTextureParams, with ResourceSubData (48) absent, which is item B working on the wire it was written for. |
||
|
|
618395529f |
[Perf] (MG_Remote/Transport): the per-op reply-wait split also lands in the exit summary, because a spawn client never prints a frame line
- The tally added beside this is only useful if somebody can read it where the problem is. The handoff says why that was hard: under spawn the client does not advance the frame window, so the per-frame P65LinkMetrics lines never print, and a SIGKILL takes the exit-time dump with it - "only the dump at exit, and SIGKILL does not get it". - LinkMetricsEnd writes the split on its own line from the one place a client that ends the ordinary way always reaches: P65LinkMetrics kind=per-op wait_replies=N by_op=<id>=<count>,... . Ids stay raw (this file includes only <cstdint> and friends, not the op table) and the reader names them: 2 ResourceCreate, 3 ResourceRespecify, 47 SetTextureParams, 48 ResourceSubData. The mapping is written down in mgl-device/README.md, which is the page the device run reads. |
||
|
|
fa8b500ba8 |
[Perf] (MG_Remote/Transport): the reply-wait metric gains a per-op tally, and the measurement says which rows actually buy the waits
- The handoff measured the total (55,428 reply waits in one Minecraft frame) and left "which rows" open: a total is the same number whether the waits are uploads, creates or parameter sets, and those three have different fixes - so knowing the split is what says whether P12 item B is the whole story or a fraction of one. - LinkMetricsBeginReply now takes the op and Window carries a per-op tally, so LinkMetricsReplyWaitsFor(op) reads back what the P65LinkMetrics line prints as `wait_replies` split by row. Same activation as everything else in that file (MOBILEGL_PIPE_STATS=1), one array in the window, and a value past the table lands on the last slot so the per-op numbers still SUM to the total. - RemoteClientControls.ReplyWaitsByOpForAnAtlasShapedLoad drives the two shapes a stitcher produces and prints the table: an atlas stitch (64 glTexSubImage2D into one level) goes from 64 reply waits to ZERO, and a texture per sprite (64 textures, 256 glTexParameteri) from 384 to 320 with the remainder all ResourceCreate 76 / ResourceRespecify 130 / SetTextureParams 130. - It prints before it asserts: the table is the point of the case, and a gate that fired first would hide the numbers on exactly the run that needs them - the red one. The assertions are the structural ones that can be wrong about the mechanism (upload rows zero, the other rows not, the per-op tally summing to the total); the numbers themselves are recorded, not gated. - The A/B is a real run: with the wire-arm arm test removed the case goes red with the full table, which is where the 64 and 384 above come from. |
||
|
|
33e02f4bcf |
[Perf] (MG_Pipe, MG_Remote/Client): on the wire arm a texture upload stops owning a reply - the wait moves from every record to the SEG_STAGE window
- MGPipeSubDataWantsItsReply is the ONE place that decides whether a resource_subdata record waits
for its answer (WireTables.cpp:370 and PipeRoute.h:328 both read it, and the comment beside it
says that a second spelling is how the client comes to wait for an answer the emitter told the
server not to bother with). The buffer half has answered "no" since P5e; the texture half now
does too, ON THE WIRE ARM ONLY.
- THE ANSWER IS DERIVABLE THERE, which is the argument - not that nobody reads it:
* The emission is already gated on the answer's content. A texture record is produced only where
FamilyIsLive(kMGPipeSubsystemTextureResources, ...) holds, and under split that gate IS
CapsMirrorInstance().ServerConsumes(bit 10) (PipeFill.cpp:1719-1740) - the same published fact
the applier's own acceptance belt reads (PipeApply.cpp:1408-1419, NoP4aConsumer). So the one
refusal this acceptance exists to carry, "no consumer" (ID-39 / D-D5 / ID-18 M3, which is why
the texture half ever asked), cannot be the answer to a record this client emitted.
* What remains is a bug class, and it is loud already: a record that names no texture, declares
no texels or carries no bytes is an MGP_TRIP_WIRE_REPORT (PipeApply.cpp:1201-1218, :2245-2252),
and a handle the applier does not hold is counted and named (ResolveResourceIn, :721-728) with
the client's own tracker preventing it from arising (PipeFill.cpp:1045-1051).
* Every answer the client can still get for free, it still gets: the emitter reads this route's
Bool exactly as before, so a record the CLIENT cancelled (the device-lost latch, an exhausted
stage) still answers "not accepted" and the level stays dirty and is retried - D-D5's safe
direction, unchanged.
- THE WAIT MOVES TO THE STAGE. A run that does not fit SEG_STAGE reclaims retired bytes and parks
on the stage doorbell (PipeWireCodec.cpp StageAllocate / ReclaimStagedBytes) - the one wait a
run-ahead client owes, and it is proportional to the WINDOW rather than to the call count. On
Minecraft 26.2's atlas frame that is the handoff's 55,428 reply waits at ~4.4 ms each: 96 MiB of
uploads through a 32 MiB window.
- THE GATE IS THE NUMBER ITSELF. LinkMetricsReplyWaits is added so a case can read back the same
`wait_replies` the per-frame P65LinkMetrics line prints - LinkMetrics.cpp:76-80 counts exactly
the records that own a reply slot. TextureUploadsStopOwningAReplyOnTheWireArm asserts that a pure
glTexSubImage2D crosses the wire and moves it by ZERO, while a resource_create row - kWaitReply,
and it always was - still moves it, so the zero cannot be "nothing was counted". Red-once run:
with the arm test removed the child exits 142.
- Monolith and every non-wire arm are untouched, and that is pinned twice: the existing
RemoteRunAhead.OnlyTheTextureHalfOfResourceSubDataWantsItsReply still reads true with no arm
installed, and the new test sits inside MOBILEGL_BUILD_DISAGGREGATED with the original expression
as its fall-through, so a pull build preprocesses to the line it had (G1).
- Verification debt, stated rather than implied: the host gate is the COUNT, not a wall clock.
In-process control sessions run lockstep (the caps peer publishes no kCapRunAheadApply), so the
verb barrier still waits there; the end-to-end wall-clock win is the device measurement the
handoff describes and it has not been taken yet.
|
||
|
|
3874b547a9 | [MG_Remote, MG_Test] (P7): band ReadPixels larger than one reply slot | ||
|
|
efc899a1b0 | [MG_Remote, MG_Test, scripts] (P7): bound the pre-auth cap and pause accepting when out of descriptors | ||
|
|
3d7cc9592b | [MG_Remote, MG_IntegrationTest, CI, scripts] (P7): add Ph fuzz arm 1 for malformed control frames | ||
|
|
eaef83b32a | [MG_Remote, MG_Test] (P7): the server binds its deferred stream link through the Transport factory | ||
|
|
8c8fc31952 |
[MG_Remote, MG_Test, ci] (P7): the token compare runs a fixed 64-byte width, an ignored window ask is logged once, a nonce that could not be minted is refused by name, and a garbage first frame while busy is MalformedHello rather than Busy
- AuthToken.h: ConstantTimeTokenMatch loops kAuthTokenCompareBytes (64) iterations with both operands read as zero past their ends - its trip count no longer tells a peer the configured token's length; a token longer than 64 is still compared to its last byte (SocketTransportTest pins the 70-byte case)
- ServerSession.cpp: an ask in Hello.linkTerms above the granted terms logs one MGLOG_W naming both sides ("the ask is ignored and the session is server-sized"); the production client asks for nothing, so a session that was never mis-configured never logs it; SessionHandshakeTest's 64 GiB case requires the line (red before: absent)
- ServerSession.cpp: a MintNonce failure sends Refuse{LinkTerms} "stream data plane unavailable: the server could not mint a data nonce" before returning the fault it is (still exit 67, still counted) instead of a bare close
- ServerMain.cpp: a first frame with the wrong magic or over 1 MiB while a session is live is Refuse{MalformedHello} "first frame is not a control frame", drained before the close so the refusal is not RST away; Busy stays the answer for a well-formed frame that is not a DataBind
- tcp_supervisor_smoke.py: garbage_first_frame_while_busy (magic XXXX on a live session -> code 8; red before: code 7)
- fatal census unchanged: 79 abort sites, 3 refusal words (the new refusals go through the enum, not a literal)
|
||
|
|
0293aebb17 |
[MG_Remote, ci] (P7): a bound session closes its data-connection source, so a later DataBind is refused by name instead of queued into a child that stopped reading, and the single-session listener closes with it
- ServerMain.cpp: RunSession takes the descriptor its data-connection source reads from (the child's hand-off end under --serve, the listener itself without it) and closes it the moment Accept returns; the source object goes with it
- the supervisor's hand-off send is MSG_DONTWAIT (FdPassing::SendFd dontWait), so a peer that stopped reading answers at once; "peer gone" on a datagram pair (ECONNREFUSED, ENOTCONN) is an answer the caller names, not a logged fault
- before: every DataBind after the bind sat in the child's receive queue with no reply, its TCP connection held open until the session ended, and past the queue limit (net.unix.max_dgram_qlen 10 on the phone; ~288 datagrams of sender buffer on this host) the supervisor blocked in sendmsg for the rest of the session - no accepts, no Busy, no reaping
- before, without --serve: the listener stayed open after the data connection bound, so a second client or a retry completed a TCP handshake nobody would accept and waited out its 20 s connect budget instead of the ECONNREFUSED it got before PH-7 (4)
- tcp_supervisor_smoke.py: data_bind_after_bound (12 extra DataBinds on a bound session each get Refuse{Authentication} "data connection could not be handed to the live session", and a fresh control connection is still Busy) and single_session_listener (a connect after `ready` is ECONNREFUSED); connect_control retries ECONNREFUSED at startup for pair() and held_session() alike
- red-once at da313010: smoke TimeoutError on the first extra's receive; probe with qlen+2 fire-and-forget DataBinds: 288 queued silently, then the supervisor stopped accepting (fresh control: connect timed out) while still alive; single-session connect after ready succeeded
|
||
|
|
20ab14daf8 | [MG_Remote, MG_Test, ci] (P7): a TCP session's data connection is bound to its authenticated control connection by a 128-bit Welcome.dataNonce instead of by arrival order, so a reordered or stale data connection is refused by name and the wire fingerprint moves with the control schema | ||
|
|
8fa1821ab2 | [MG_Remote, MG_Test, ci] (P7): one constant-time token comparison serves both server policy sites, a configured token shorter than 16 bytes refuses the listen by name, and CI builds flatc so the TCP supervisor smoke runs and must pass | ||
|
|
f141918aa1 |
[scripts, MG_Remote, MG_Impl, MG_Test] (P7): the fatal census gate widened to the split server image
`scripts/ci/fatal_census.py` scanned `MG_Remote/` and saw four abort sites. CONTRACT-P6
§12.1 already says what the server image is - "P6's server image is the whole
libMobileGL.so" - and the server process was executing seventy-nine more deaths the gate
could not see: 15 in DirectGLES.cpp, 12 in Managers.cpp, 11 in VkBufferManager.cpp, 7 in
PipeRoute.cpp. SCAN_ROOTS is now MG_Remote / MG_Pipe / MG_Backend / MG_Impl/Pipe /
MG_State, `.inc` joins `.cpp`/`.h` (the Magma wire arms and the generated wire records are
.inc files, and 20 sites are reachable only through one), and MG_Benchmark /
MG_IntegrationTest join MG_Test in the exclusion.
Four rules now, where there were two:
1. an abort names its family within twelve lines, OR it is argued into FUNNEL_FILES /
the new FUNNEL_SITES table. FUNNEL_SITES exempts an individual funnel inside a file
that has ordinary named deaths around it, and only when one of the file's anchors
appears in the same window - so the exemption follows the funnel if it moves, expires
if it is deleted, and does not cover an unrelated abort added to the same file. Two
entries, one line of reason each: PipeApply.cpp's MGP_TRIP_WIRE_REPORT and
PipeWire.inc's MGPipeWireProtocolFatal.
2. unchanged: a WireLogFatal / SessionFail call carries a `Fatal{` in its string.
3. NEW: every family word found in the perimeter has a row in FatalFamilies.def. Eleven
words did not, and now do (see below).
4. NEW: every `Refuse{Word}` names a RefuseCode enumerator. This is what catches
`Refuse{AuthenticationRequired}` (SocketTransport.cpp:161), a word protocol.fbs has
never had - PH-7 (2). Renamed to the enum's own `Authentication`, and
SocketTransportTest pins the spelling to `EnumNameRefuseCode` so the two cannot
drift again.
`abort_sites` was recorded and never compared. It is now a DOWNWARD ratchet with a
per-file map beside it, so "83 became 84" reports as "VkBufferManager.cpp went 11 -> 12,
here are its lines"; fewer sites - the direction `@P7` retirement moves the number - asks
for a re-baseline rather than blocking one.
Eleven new `.def` rows for the eleven rowless words. Ten are ProtocolCorruption for the
reason that bucket already states. ResourceUnavailable is ServerCrashed, not SegmentMismatch:
SegmentMismatch means a shared SEGMENT was the wrong size, and telling a peer that when a
VkBuffer allocation failed would be a lie about whose geometry was wrong. A twelfth word was
collapsed rather than given a row - `InitialBytesNotCarried` (PipeFill.cpp) said the same
thing as `UncarriedInitialBytes` (WireTables.cpp) about the same record, and two words for
one family is the drift a6 censused.
Red-once, four ways: an unnamed `std::abort()` appended under MG_State reds naming the file
and the line; deleting the `Fatal{` from PipeRoute.cpp's ResidualBlockSize reds; putting
`Refuse{AuthenticationRequired}` back reds rule 4; a typo'd family word reds rule 3 and the
vocabulary ratchet together.
|
||
|
|
fe28bdb5f3 | [MG_Remote, MG_Pipe, MG_Test, CI, tools, docs] (P6.5): the first wave completed and reviewed - archive v2, publication boundary, half-close, AcceptPair/role fixes, device re-validation | ||
|
|
018fab0f5b | [WIP] (MG_Remote): implement P6.5 TCP and stream transport | ||
|
|
872370ba9d |
[docs, MG_Remote] (P6->P6.5): the end state is TCP across machines, and the transport stack has two axes
The end state is re-ruled: client and server may sit on different machines, OSes and architectures, joined over TCP. The AVF pVM / AF_VSOCK end state the P6 end-state review assumed is withdrawn (retraction at its head, form D added to its table, the vsock probe and both external unknowns struck). The transport stack becomes two independently selectable axes, negotiated in the handshake and mixable: control plane (ITransport: fork / unix:<path> / tcp://host:port) and data plane (ILink: ShmLink / StreamLink). On one machine control over TCP with data over shared segments is a legal pair; ShmLink brings its own AF_UNIX aux rendezvous for SCM_RIGHTS, so ITransport narrows to control frames. `auto` is decided by a successful segment delivery, never by inspecting addresses, and a fall-back is named. Nothing above Transport/ may branch on the link kind. ARCHITECTURE.md gains 11.9 for this; CONTRACT-P6.md gains 10.3. ROADMAP: P6.5 and Ph get rows (both were referenced by the contract and absent from the stage table); the IPC track becomes P6 -> P6.5 -> Ph -> P12 -> P9 -> P10 -> P11(same-machine arm). Every later row is reconciled against the tree: items already landed are struck (client vertex arrays, client-minted query handles, credit default 1, SEG_EVENT overflow mechanism), the unreachable P7 gate (`nm | grep glslang` on a server image that IS libMobileGL.so) is replaced by the 184-symbol ratchet share, fixed-width wire form moves from P7 to P6.5, P12 is cut down per the review (server-owned display, no Surface passing), and the debt table's stale claims are corrected: class C is down to two slots, not 15 and not 511; PACK-PBO and the 2 MiB reply cap are P9's, not P6's. Verified against the tree and recorded honestly: the contract's three hs rows (LinkTerms, Refuse, the fingerprint split) never landed - protocol.fbs has neither - so CURRENT_STAGE_PROGRESS.md's claim that S6 ran red is withdrawn and the rows move to P6.5; the 184-symbol ratchet gate does not exist in CI; the server process never initialises PipeStats, so gate 8's `srv=0 srvpark=0` is a structural zero (fix in progress in the tree, not in this commit). check_doc_citations.py: 159 citations, one pre-existing basename ambiguity. |
||
|
|
90f9a79058 |
[MG_Remote] (P6/hs): the build stamp that was silently empty, and two asymmetries the handshake never checked
CONTRACT-P6 4.2 and 4.3. Three defects, all of them things the handshake
believed it was checking and was not.
1. THE BUILD STAMP WAS "" AND NOTHING SAID SO. CMakeLists took the git hash
through an execute_process with NO RESULT_VARIABLE, so a failed
`git rev-parse` yielded the empty string. Verified live rather than
inferred: this worktree's `.git` is a file holding a WINDOWS path, WSL's
git answers `fatal: not a git repository`, result 128 - and every WSL build
mixed "" into its ABI fingerprint. Two builds from different commits then
agreed exactly, which is the one thing the stamp exists to prevent. The
already-failing SessionHandshakeTest was reporting this the whole time.
The fix may NOT touch GIT_COMMIT_HASH_SHORT: it is baked into
glGetString(GL_VERSION), so redefining it changes an application-visible GL
string and the pull build's .rodata - a behaviour leak and a G1 break at
once. MGGitHash.h.in gains a SECOND pair instead, consumed only by the
fingerprint, plus -DMOBILEGL_BUILD_STAMP for trees git cannot read (CI has
${GITHUB_SHA} in hand). A stampless build is now a configure-time WARNING
naming the consequence, not silence.
MOBILEGL_BUILD_STAMP_PRESENT IS MIXED AS ITS OWN FIELD. "this build cannot
name its commit" and "this build's commit is the empty string" are different
claims, and collapsing them is how every stampless build in the world agreed
with every other. The test asserts the flag moves the fingerprint in EITHER
build state, and the `BuildStamp = ""` perturbation is now guarded - in a
stampless build production's value already IS "", so that assertion was
passing while proving nothing.
2. THE CLIENT NEVER READ abiMajor/abiMinor. Hello has carried them since P5
and the SERVER checks them; the client compared only the fingerprint. The
fingerprint mixes AbiVersion so a mismatch would fire anyway - but as
"struct shapes", which sends the reader looking for a layout change that is
not there. Under spawn the peer is a SEPARATE IMAGE the operator chose with
MOBILEGL_IPC_SERVER_PATH, so this is exactly the failure that needs its own
sentence. Both handshake sites call one helper; there are two sites, and
only one of them was reachable before `sm`, which is how the gap survived.
3. Hello::pid WAS 0 AND Welcome::serverPid ECHOED IT. So both were always 0 and
neither named anything, while §9.5's arm-proof gate wants exactly the child
pid. Both now carry real ids, and the pair is the cheapest arm proof in the
protocol: under spawn they MUST differ, and a same-process session cannot
produce two different values. The client keeps what the server STATED -
local launcher knowledge is a different fact from one that crossed the wire,
and only the second can be asserted from the server's own mouth.
R-16 both ways: revert serverPid to echoing hello->pid() and the new test reds
on both halves, naming 4.3. G1 re-measured after the CMake and header changes:
0 symbol diff, .text unchanged at 0xa52203.
Unit suite goes from two long-standing failures to one - the remaining
PipeCatalogue.FrontendNeverTakesAnApplierAddress is an environmental path
problem in this tree, unrelated. All six OpenRA arms still green.
|
||
|
|
8b13d8a84c |
[MG_Remote, MG_Impl] (P6/dl): the device-lost latch, taken from a descriptor and never from a deadline
Before this, CONTRACT-P6 5.3's own sentence held: "a dead server leaves the
client silently DECLINE-ing every verb forever with glGetError clean".
FatalCode::DeviceLost had zero producers and zero consumers and
glGetGraphicsResetStatus returned GL_NO_ERROR unconditionally, so an
application had no way to learn that the process holding its GPU objects was
gone. Measured cost of that: a server killed on its first Clear left the
client waiting out the full 120 s barrier and reporting Fatal{BarrierTimeout}
- the wrong diagnosis, two minutes late. It now notices in 0.01 s.
THE MACHINERY WAS ALREADY THERE. Doorbell::Wait breaks out on Dead(),
SessionProducer::Park turns that into SessionWait::ShutDown, and ClientSession
already had the named DECLINED path. What did not exist was the death FACT:
D5c's finding is that the client parks on clientBell[0] and rings ITSELF
through clientBell[1], holding both ends, so EOF can never arrive on the one
descriptor it watches. A bell that can wake itself cannot hear a death.
SetDeathWitness(fd) fixes that with no new descriptor: the control socket's
far end is held only by the server. The subtlety is that the bell must watch a
socket it must NEVER read - the transport's reassembler owns those bytes and a
control reply is not a death. So the witness enters the poll set with
`events = POLLRDHUP` and nothing else: POLLHUP/POLLERR/POLLNVAL arrive in
revents unrequested, so the hangup is seen while POLLIN never is. Nothing is
recv'd from it and its ownership stays with SocketTransport.
Dead() AND PeerHungUp() ARE TWO FACTS. Park sets m_dead for POLLERR, POLLNVAL
and unrecognised revents too - each a fault in THIS process's descriptor - and
CondVarDoorbell::Kill() sets it on every orderly Stop(). A latch taken from
Dead() would report a lost GPU for a local fd bug and a context reset for
every clean exit, so only PeerHungUp() arms it.
GL_UNKNOWN_CONTEXT_RESET, not GUILTY or INNOCENT: those assign blame, and the
server died for a reason that never crossed the wire.
D5b, in the control-reply wait, which had the right comment and no code: EOF
on the control stream is device loss; a TIMEOUT with the socket still open is
"ALIVE BUT SILENT" and says so by name. 5.4 forbids calling a frozen or merely
slow peer dead, and under P5e run-ahead a server one frame behind is the
INTENDED steady state - a latch armed by a deadline would fire on a healthy
session under load.
MOBILEGL_IPC_RESPAWN is now a NAMED REFUSAL rather than a variable nothing
reads. The latch is one-way on purpose: every handle this client minted names
an object inside the server process, so recovery means re-pushing the world.
THREE TEST NAMES ADDED, none renamed (G14), and R-16 run both ways:
* remove the witness -> AKilledServerLatchesDeviceLostFromTheHangupAndNot-
ADeadline goes red, takes 2 s instead of 0.01 s, and names D5c.
* collapse PeerHungUp() into Dead() -> ADeadBellIsNotAlwaysAHungUpPeer goes
red naming the local-fd-bug consequence.
* AnOrderlyStopIsNotADeviceLoss stays GREEN under that collapse, and its
comment now says so. The first draft claimed it was the control for it;
running the falsification proved otherwise - under spawn the self bell is a
SocketDoorbell and Stop() never kills one. The comment was corrected rather
than the claim quietly kept.
G1 RE-MEASURED, NOT ASSUMED: 0 symbol diff against the c6 baseline and .text
unchanged at 0xa52203. That measurement earned its keep - it caught a 16-byte
drift that had nothing to do with dl: P6's O_APPEND log-sink fix was compiling
into the PULL build, where a second process can never exist. Now guarded, and
.text is back to the baseline exactly.
Full unit suite: 2350 tests, the same two pre-existing environmental failures
and no others. All six OpenRA arms (2 backends x monolith/inproc/spawn) green.
|
||
|
|
7c8120cc49 |
[MG_Remote, android-plugin] (P6): a real phone retraces a real trace across two processes
OpenRA.DirectGLES, transport=spawn, on a Redmi: ssim=1.000000,
mismatchPixels=0, zero Fatal{ lines, three runs for three. The server is a
process the client launched out of the APK's lib/arm64/, and the client's own
log names its pid.
Four things stood between the desktop green and the device green, and every
one of them was invisible on desktop for the same reason: a build tree is not
a deployment, and Linux is not Android.
1. THE SERVER COULD NOT LINK. It links libMobileGL.so and is EXEC'd, not
dlopen'd, so it gets a fresh linker namespace with the DEFAULT search path
- and an APK's lib/<abi>/ is not on it. `CANNOT LINK EXECUTABLE ...:
library "libMobileGL.so" not found: needed by main executable`, before a
single line of MobileGL ran. On desktop CMake had been giving the target an
absolute build-tree RPATH, so the WSL server resolved its dependency
through a path no phone has. $ORIGIN is asked for via BUILD_RPATH (a raw
-Wl,-rpath,$ORIGIN reaches the generator unquoted and the SHELL eats it -
the device linker said `normalize_path - invalid input: "$"`), and because
the NDK drops RPATH from the link anyway, the LAUNCHER states
LD_LIBRARY_PATH: it is the one component that has already resolved where
the image lives, and a pair that is not co-located is not a pair.
2. THE RENDEZVOUS NEEDED A WRITABLE DIRECTORY. An Android app has no /tmp, so
bind() in the child had nowhere to put the node and the client refused by
name 20 s later. AF_UNIX endpoints now accept the ABSTRACT namespace with a
leading '@' - the systemd/D-Bus convention - which has no filesystem
presence at all: no directory, no mode bits, and no stale node to unlink
after a crash, which is the bind() failure Listen() has been working
around. Its reach is the network namespace, which for an app is the app.
StartSpawned mints '@' names; an explicit path still works everywhere, and
that is what P12's "connect to something somebody else is listening on"
will use.
3. PBUFFER WAS A DirectGLES-ONLY KNOB. Twice over: trace-replay-ci.sh gated
`--ez use_pbuffer true` on the backend name, and the JNI demanded a native
window for DirectVulkan unconditionally. A surface SHAPE is not a backend
property, and the coupling mattered because the Android GLWS makes a window
surface whenever gNativeWindow is non-null - so handing Magma a window was
the same as asking for one. Until P12 a spawned server cannot be given a
window at all: an ANativeWindow* is a pointer into the CLIENT's process and
SetWindowHandle is refused on the way across with Fatal{UnmigratedSurface,
"AndroidNativeWindow@P12"} (Rule H). Which is exactly what happened, one op
after the server had come up green.
4. THE RUNNER LOOKED IN ONE OF THE TWO OUTPUT DIRECTORIES. assembleTraceRelease
writes apk/trace/release; CI has staged the same artifact under
apk/trace/debug. A tree that has built locally has both, and picking the
wrong one retraces last month's APK and says nothing about it. Newest of
either, now.
MEASURED AND NOT FIXED: DirectVulkan on this device diverges under BOTH split
transports and does so nondeterministically - inproc scored ssim=0.988998 /
9836 px, spawn scored 0.976494 / 14658 px and then 0.988998 / 9836 px on a
repeat, while monolith with the same pbuffer surface scores 1.000000 / 0 px.
The two transports draw from the same set of outcomes, so this is the split
path on Magma against a real GPU, not the process boundary: P6 neither caused
it nor can fix it. DirectVulkan spawn IS green on desktop (lavapipe, 1.000000).
G14: one test name added, AnAbstractEndpointNeedsNoWritableDirectoryAndLeaves-
NoFile, which pins both halves - the handshake crosses an abstract name, and
nothing is left on the filesystem. A '@' mistaken for a path would have bound
a file of that literal name, passed here, and failed on the one platform this
exists for.
|
||
|
|
7e244b3775 |
[MG_Remote] (P6/cp): a real trace retraces across two processes, pixel-for-pixel
OpenRA.DirectGLES.SPAWN is green: ssim=1.000000, mismatchPixels=0, zero
Fatal{ lines, against a server in a process the client started by name. The
inproc SPLIT arm stays beside it as the control, so "did the transport break
it" and "did the phase break it" remain separately answerable.
Five defects stood between the handshake and that picture. Each one was
silent in its own way, and none of them looked like what it was.
1. THE DOORBELL COULD NOT WAKE ITSELF. SocketDoorbell's one-fd form sends on
the end it polls, so a write is delivered to the PEER - correct for waking
the other process, useless for waking a thread in this one. The server's
control pump posts into the apply thread's mailbox and must ring the bell
that thread parks on. Symptom: every eglInitialize crossed, landed in the
mailbox, and timed out at 5 s. The two-fd form (park on [0], notify on [1])
is restored here, and the bell handover goes from two descriptors to THREE
- the client needs the same split for the same reason. Seven offers now.
2. InitCapabilities HAD NO WIRE KIND. CONTRACT-P6 §2 predicted this exactly
("the request has a live client caller and the codec refuses to encode
it"). SurfaceOpKind gains 11, append-only; the decode range's upper bound
moves with it, or the new tag reads as out-of-range and surfaces as
Fatal{ProtocolCorruption} rather than the missing row it is. The op's name
loses its InprocOnly suffix, which stopped being true here.
3. THE CHILD LOADED NO CONFIG AT ALL. mobilegl_server_main calls
InitServerRoleForSpawn directly, and MG_ConfigLoader::Init is a step of
MobileGL::Initialize - so every MG_Config value in the server was its
static default. MOBILEGL_BACKEND_TYPE survives the launcher's scrub on
purpose and was read by nobody: a DirectVulkan session would have brought
up a DirectGLES server and then disagreed with its client about which
backend the capability mask describes.
4. AND THEREFORE Transport READ AS Monolith, which CONTRACT-P6 §3.1 (D1)
forbids in as many words. a6 measured the cost: 226 live lines test
`Transport != Monolith` to select the SERVER arm, 148 in MG_Backend.
Monolith flips all of them to frontend glue in the one process with no
frontend. Observed precisely as predicted - UnitTexturesByHandle() answered
false, Clear took the legacy arm that dereferences borrowed FRONTEND slot
pointers, and the apply thread took a SIGSEGV inside SyncNeccessaryTextures
on the first Clear record of the trace. It is pinned in-process rather than
through the environment because the environment is where it must NOT be:
LooksLikeAClientEnvironment exits 65 on a surviving MOBILEGL_TRANSPORT, and
that check is anti-recursion catch (b). Dial, not Transport, is that axis.
5. THE LOG WAS CORRUPT, NOT MERELY INTERLEAVED. Two processes opened the same
MOBILEGL_LOG_FILE_PATH with "w" and each carried its OWN file offset, so
they overwrote each other's bytes - half-lines like `ote client: CapsMirror
generation 1 adopted`, and the server's whole startup block vanishing under
the client's later writes. This cost real diagnosis time twice, because a
missing line read as a code path not taken. O_APPEND now; the client still
truncates once so a fresh run is not a pile of old ones, and the server
never does - it would delete the very startup lines a failed session needs.
run_trace_case.cmake's Fatal{ census reads exactly this file, so it was
also reading a lie.
The retrace runner's split assertions become transport-aware: one row per
transport it admits, and a transport with no row is a FATAL rather than a
fall-through, because the fall-through is the accident the block exists to
prevent. The spawn row carries a SECOND sentence and it is the stronger half
- ConfigLoader's marker proves the value RESOLVED, while `spawn ARMED - the
server role runs in pid N` is written only after the launch, the connect and
the handshake all succeeded, and carries a pid that is not this process's.
inproc resolves its own marker while running the server role on a thread
here; nothing on that path can write this one.
G14: one test name added (TheEleventhFramedOpIsInitCapabilitiesAndItCrosses-
TheWire), none removed. The ten-op table keeps its name and its scope rather
than being renamed to eleven.
Two unit failures predate this change and are environmental in this tree:
PipeCatalogue.FrontendNeverTakesAnApplierAddress (the frontend source root
resolves empty) and SessionHandshakeTest.TheAbiFingerprintChangesWhenAnyOf-
ItsInputsDoes (GIT_COMMIT_HASH_SHORT is empty here, which a6 already recorded
as red in this configuration).
|
||
|
|
11807ac7d0 |
[MG_Remote] (P6/sm): two INDEPENDENT processes, joined by a name, sharing memory
The cross-process session is up. A server process is started on its own with an
endpoint and nothing else, it listens; a client started separately connects to
that name; Hello crosses, Welcome comes back with four SegmentRefs, six
descriptors cross by SCM_RIGHTS, and the client maps memory another process
created. ClientSession::Active() is non-null at the end of it, which is the flag
that arms the thirty-seven wire emitters.
INDEPENDENT, NOT FORKED - and that is a correction to this package's first
shape. The version committed an hour ago forked the server and handed it fds
3..6. It worked, and it was less code, but inheritance can only ever produce a
server that is a CHILD OF ITS CLIENT, which is the one arrangement the end state
cannot use: there the server is a standing application and the client is
somewhere else entirely, possibly on another kernel. CONTRACT-P6 §3.1 called
this Dial == Connect and filed it as a later shape. It is the primary one, and
fork/execve survives only as a way to START a process - every descriptor it used
to carry is gone.
What that costs and buys, concretely: the rendezvous becomes a NAME, so
SocketTransport grows Listen/AcceptPair/ConnectTo, with a bounded connect retry
that is the CLIENT's (a launcher that waited for the server to be ready would be
coupling the two again by the back door). The bells stop being inherited fds and
are created by the server and handed over with the segments - slots 4 and 5 of
six SCM_RIGHTS offers. And the client connects TWICE to the same path: an fd
offer is a sendmsg whose ancillary data rides with specific bytes, so it may not
share a socket with the framed control stream, or the frame reassembler and the
descriptor receiver race for the same bytes.
The pieces behind it:
- SessionSegments::AdoptFromDescriptors, which is the same Adopt+Map loop
AttachInProcess has run on every inproc session since P5. That was the point
of writing the inproc attach as dup+Adopt+Map rather than as an alias: only
the origin of the descriptor is new.
- ServerSession::SetExternalDoorbells, the narrow fix for a6 row A7-1 - the
§3.9 accessors abort for every role but InProcess, so a spawn apply thread
died before popping its first record. lk deletes them; until then an injected
pair is honest about where the bells come from.
- MG_Backend::InitServerRoleForSpawn: steps 1 and 2 of the split bring-up
extracted so the server process runs the SAME code the inproc server role
runs. The capability bits are a statement about the server's own backend and
a second copy would diverge the first time a backend registers a slot.
- ClientSession::FinishStartup: everything below the attach - producer, reply
pool, event ring, segment table, encoder, first CapsSnapshot, emitter tables -
shared verbatim between inproc and spawn. Its one parameter is not a
transport question: under inproc the client starts the apply thread because
there is no server-side main; under spawn there is one, and it started the
thread before it sent the first descriptor.
G1 CAUGHT A REAL LEAK on the way: InitServerRoleForSpawn was guarded on the
inside, so the pull build still DEFINED the symbol - 2 added, 1 removed, .text
+16 bytes. The whole function and its declaration are now inside the guard.
"It returns false there" is not the same as "it is not there", and the gate is
what said so.
Gates: ServerSpawnTest 4/4; unit suite 2344 with the same two pre-existing
environmental failures; G1 pull 0 added / 0 removed, .text 10822147 unchanged.
|
||
|
|
5f846d96e3 |
[MG_Remote] (P6/so): SocketTransport, green over a real socketpair before any second process
The first P6 code that runs. ITransport over a connected AF_UNIX SOCK_STREAM socket, with SCM_RIGHTS descriptor passing on a dedicated SOCK_DGRAM aux pair. No shared memory: this is the CONTROL plane, and the hot path bypasses ITransport entirely by its own header's account. What the socket carries is the handshake, surface ops, resync, aux requests and fatals - and, once sm exists, the four segment descriptors that let the peer map the data plane. It moves the keys, not the bytes. CONTRACT-P6 §8 and the spawn plan both require this green over a plain socketpair with two threads BEFORE a child is ever forked, for the reason SessionRings.h gives about the attach half: a transport first exercised on the day the process appears is a transport whose bugs all arrive at once, on the hardest day to debug them. That requirement is now met. It reinvents nothing. Framing.h already had AppendFrame and a real byte-stream reassembler with exactly the BUFFER_TOO_SMALL keep-the-message semantics ITransport specifies - written for this and never before driven by a socket. FdPassing.h already had SCM_RIGHTS over a dedicated datagram pair. What is new is ~250 lines joining them, plus the platform refusal: Windows has no SCM_RIGHTS and ShmSegment::Adopt is POSIX-only, so CreatePair refuses BY NAME at construction rather than handing back a transport that fails every later call. SOCK_STREAM rather than SOCK_SEQPACKET, deliberately: Doorbell.h:390-398 already chose a stream socket so a closed peer surfaces as POLLIN|POLLHUP with recv()==0 rather than as a hang, and the control plane may not disagree with the bell about what death looks like. Framing.h is what makes a stream safe - the message boundary is in the payload, not in the socket. TESTS: the same nine ITransport-contract cases InProcessTransportTest runs, copied rather than shared. Folding both suites into one parameterised fixture would rename the sixteen existing cases and G14 says names only grow; and the two suites do not share a handle anyway, since InProcessTransport's tests reach into condvar doorbells this class does not have. Nine names added, zero removed. The seven Doorbell* cases are deliberately absent: they exercise CondVarDoorbell, not ITransport. Plus three a stream has and a message queue cannot: a 256 KiB payload that provably crosses in several reads; a garbage stream that must latch failed and must NOT resync onto a boundary it invented; and adopting an already-connected fd with no aux channel, which answers UNSUPPORTED to ShareFd rather than pretending - rule G, a transport declares what it supports. Red-once, both run: drop the message on BUFFER_TOO_SMALL and six cases go red; make the framing latch recoverable and AGarbageStreamIsLatchedFailed fails by name. Gates: 12/12 in 677 ms. Unit suite 2340 tests, up 12, with the same two pre-existing environmental failures. G1 pull 0 added / 0 removed, .text size 10822147 unchanged, and zero SocketTransport symbols in the pull build. One note for the VM track this eventually serves: swapping AF_UNIX for AF_VSOCK is one line here and nothing at all in Framing. It is the shared-memory half that does not cross a kernel. |
||
|
|
db62d674d0 |
[MG_Remote] (P6/lk): regroup RingControl by writer, and make the asserts that pin it falsifiable
The shared control page both roles map had a mixed-writer cache line:
{appliedSeq, submittedSeq, retiredSeq, completedFrameSerial, presentAckSerial}
sat together, but submittedSeq is producer-written (Ring.h's "THE FIVE
WATERMARKS") while the other four are consumer-written. That is why no
single-writer progress block could be aliased over it, and it is what this
package fixes: submittedSeq moves up beside cmdHead, the other thing the
producer writes, and the four that remain become a named LinkProgress member -
the struct c6 declared in ILink.h - so a data plane with no shared page can own
the same four values and every wait predicate in SessionRings.h keeps ONE
implementation at source level.
The two event flags deliberately stay in the doorbell group. Promoting them
would put the client's per-drain exchange(0) on eventRingFull (EventRing.h:200)
onto the line the client's own spin predicate reads 1e5-1e6 times a frame.
73 call sites rewritten, 15 in the library and 58 in tests. The rewrite is
receiver-scoped rather than field-scoped, because SplitRuntimePeek.h:77 has its
own struct with a field of the same name and SplitRuntimePeek.cpp:146 has both
on one line - a field-scoped rename would have silently broken the peek struct
that MagmaRunAheadScenario reads nine times.
THE ASSERTS ARE THE PART THAT NEEDED A SECOND PASS. The first revision shipped
four aggregate asserts and an adversarial review broke them by counterexample:
swapping retiredSeq and completedFrameSerial inside LinkProgress compiled
clean, passed all four, and passed the runtime layout test too - because
appliedSeq stays at offset 0. Three wire-visible offsets on a shared page were
pinned by nothing. Nothing else in the tree covers it either: the ABI
fingerprint takes no input from this page, and sizeof(RingControl) is 4096
whatever the members do, because alignas(4096) makes it so.
So the block is now per-field as CONTRACT-P6 §8.3 actually asked, plus a
standard-layout precondition, a tail bound that alignas cannot satisfy for us,
and the LinkEventFlags mirror. Each was run red: the field swap reds the
per-field asserts by name, a fifth member inside LinkProgress reds sizeof.
One of them cannot go red, and Ring.h now says so rather than implying
otherwise. A SMALL member inserted between Progress and serverEpoch lands at
offset 288, inside Progress's own line, and serverEpoch's alignas(64) absorbs
up to 32 bytes of it - so both line asserts still pass. The review's proposed
fix does not catch it either; I tried it. A gate nobody can falsify is worse
than no gate, so the limit is written down beside the asserts.
Also: RingTest.ControlPageLayoutIsTheSharedContract said nothing about the one
field this package moved, and now asserts the producer/consumer separation and
the event flags' exclusion itself rather than borrowing the claim from Ring.h.
Ring.h grew 30 lines, so the Ring.h:NN citations this package and c6 authored
are re-anchored to named blocks instead of line numbers. ARCHITECTURE.md §11.2
is updated for the new layout, and a factual error introduced in the previous
revision of that paragraph - submittedSeq annotated as releasing staging, which
retiredSeq does - is corrected.
Gates: pull 0 added / 0 removed, .text size 10822147 unchanged; split 0/0,
.text size 11681395 unchanged; unit suite 2326/2328 with the same two
pre-existing environmental failures as the baseline. The five edits behind
MGITEST_SPLIT_RUNTIME_PEEK were proven compiled by a red-once, not assumed.
Note on the .text claim: that is a SIZE, not the bytes - field offsets moved,
so immediates moved with them. And this page's layout IS observable across
binaries; what is unobservable is behaviour within one binary. The contract
already books the layout change deliberately, before spawn exists to care.
|
||
|
|
99649a9087 |
[MG_Remote] (P6/lk): the pre-flight itemisation, and a correction to c6's regroup design
CONTRACT-P6 §8.5 requires lk to publish its blast radius before it starts, because an earlier draft's ratchet baseline was wrong by a factor of three and a ratchet installed on the wrong number is red on its first CI run for reasons that have nothing to do with a leak. The itemisation is mechanical, per file, with the grep spelling stated: 250 seam-identifier references outside Transport/ across 23 files (127 production, 123 test), and 44 member-access sites that the regroup itself must rewrite, across 9 files - 17 of them in MG_IntegrationTest, which is why lk builds with the integration tests ON. The earlier build did not, and a missed edit there would have been green. THE CORRECTION. Measuring that radius turned up a mistake in c6, landed two commits ago. §8.3 and ILink.h both said lk should promote eventRingFull and eventDropped into the watermark group and called the result single-writer. eventRingFull has TWO writers: the server latches it with store(1) when SEG_EVENT fills (EventRing.h:140) and the CLIENT clears it with an exchange(0) on EVERY DRAIN (:200). Promoting it would put a per-drain read-modify-write by the client on the same cache line as appliedSeq - which the server release-stores every 64 records and the client's spin predicate reads 1e5-1e6 times a frame. That is new false sharing on the hottest line in the system, and it would land in exactly the numbers this package's gate has to hold constant. It is also the same class of error the earlier draft made with submittedSeq, one field over. So LinkProgress is now the four consumer-written watermarks and nothing else - genuinely single-writer - and the two flags stay in the doorbell/generation group, where mixed writers already live beside consumerParked/producerParked, crossing the seam through a separate EventFlags() accessor on its own line. The contract records the correction and its reasoning rather than quietly restating the rule. ALSO CORRECTED, from running the baseline suite: a6's link-experiment report said the silently-empty build stamp had "no log and no gate to say so". That is false. SessionHandshakeTest.cpp:140 asserts BuildStamp[0] != '\0' and :180 asserts that perturbing it changes the fingerprint, and BOTH GO RED in this configuration right now - the worktree's .git holds a Windows path WSL's git cannot resolve, so GIT_COMMIT_HASH_SHORT is "". The finding is therefore stronger, not weaker: the gate that catches it already exists and is already red. What is silent is the BUILD, and hs owes a named configure-time failure rather than a new gate. Baseline for lk, recorded: 3878 tests, two failures, both environmental and understood - the fingerprint one above, and PipeCatalogue.FrontendNeverTakes- AnApplierAddress, which walks up from cwd looking for MobileGL/MG_Impl and so fails for any out-of-source build directory. No behaviour change. The headers are still included by nothing. |
||
|
|
b6293957e9 |
[MG_Remote] (P6/c6): the normative contract, and the data-plane seam declared but not wired
c6 is P6's contract package. It lands CONTRACT-P6.md plus two headers and changes nothing: no caller moves, no byte moves on any wire. THE CONTRACT supersedes P6-CONTRACT-DRAFT.md, which was written before its own audit and is wrong in four places - the ABI row, "the latch is implemented unchanged", "cp is transport only", and "a transport swap and nothing else". a6 measured all four. It carries 21 rulings, of which the load-bearing one is D1: the spawned child keeps Transport == Spawn and anti-recursion becomes a separate Dial axis. ARCHITECTURE.md:488's literal instruction is amended, because 226 live lines test `Transport != Monolith` to select the SERVER arm and forcing Monolith flips 148 MG_Backend lines to frontend glue in the one process that has no frontend. That ruling can be stated normatively without changing behaviour because of something a6 found and the contract exploits: TransportMode::Spawn is UNREACHABLE at this head. ConfigLoader.cpp:344 is the only writer for the string "spawn" and it writes Monolith. So every rule conditioned on Transport == Spawn constrains a state no process can enter - and the same fact is why ConfigLoader's named refusal cannot currently falsify itself, which is sm's to fix and §9.5's arm-proof gate to catch. THE SEAM. ITransport says in its own header that the hot path bypasses it; below it there was no interface of any kind, because Ring.h, ReplySlot.h and EventRing.h each take a raw void* plus a capacity. ILink.h is that missing thing. It is setup-time, not hot-path: no method is called per record or inside a spin predicate, because Doorbell::Wait's predicate runs 1e5-1e6 times a frame reading watermarks straight off the page, and because the only lane that could measure a virtual there has four times the dispersion of the effect - a gate that can only print "within noise" is ID-124 one level down. StreamLink.h is declared and implemented by nobody, and that is its job: ILink's adequacy is a claim only a compile can settle. Writing it forced three methods the brief did not name - ResolveSpan, FlushProgress, and the reply pair - so it has already earned its keep. Delete one ILink method it declares and the build goes red. Deliberately NOT here, each for a reason the contract records: Watermark::Reached is not redefined (SessionRings.h owns it); Ring.cpp's reserve/pop bodies are not relocated into a header that would then need the logger; RingControl is not regrouped, because that rename is 109 production sites with five behind a non-Android #if, so it is lk's with a published itemisation; and the schema rows are specified but left to hs, because deprecating SurfaceReply.defaultFb while appending `result` changes CreateSurfaceReply's arity and the existing six-argument call would silently bind its defaultFb 0 to result. INERTNESS IS STRUCTURAL, not measured. Zero object dependency files reference either header, zero build rules do, and nothing in the tree includes them. Verified anyway on both flavours: pull and split each have 0 symbols added, 0 removed, and .text byte-identical (10822147 / 11681395). check_doc_citations resolves all 61 citations. |
||
|
|
68b55ea371 |
[MG_Remote] (Disaggregated): the inproc barrier watchdog is a deadlock detector, so its budget leaves room for a leg
lavapipe/llvmpipe stall the apply thread for tens of seconds inside texture-handle
registration (vkCreateImageView -> lvp_CreateImageView -> llvmpipe_register_texture,
a JIT/futex path): one measured stall is 29.3-39.7 s and a whole inproc retrace run
accumulates ~44.6 s of it. At a 30 s bound those runs red intermittently on
Fatal{BarrierTimeout, "ResourceRespecify"} while the same stall under monolith is
merely slow - monolith has no cap on that path. Since the monolith shares the stall,
the bound, not the workload, was the defect.
kBarrierTimeoutMs becomes 120000, >3x the largest observed stall, which still leaves
the watchdog job: a record that is never retired still terminates the run and the
Fatal still names the seq. The comment now says deadlock detector rather than
performance gate and carries the measurement.
The fake-clock case that pins the budget moves with it: FenceWaitBudgetHonorsGl-
TimeoutAndFiniteTransportChunks expected 30000/30001/90000 and now expects
120000/120001/180000, so the assertion still pins the value instead of drifting.
Texts that wrote the old number down are updated: ARCHITECTURE 17.1, CONTRACT-P5E
2.4 (which now names kBarrierTimeoutMs rather than the literal), and the two
bring-up comments in PipeRoute.h / WireTables.cpp that dated a self-wait to
"thirty seconds".
|
||
|
|
a3466e6f4c |
[MG_Remote] (Disaggregated): P5e ra - SEG_EVENT stops aborting and starts flow-controlling
- CONTRACT-P5E §2.6. Fatal{EventRingOverflow}'s premise was "the client drains at
every verb, so a full 256 KiB ring is a burst no workload has". A run-ahead client
drains at its waits, which on the steady path are one present apart, so a full ring
is an ordinary backlog and aborting on it would turn the reverse channel's capacity
into a workload limit.
- The producer blocks instead (ReserveEventOrBlock, one entry point for all four
event kinds): publish what is in the ring, ring the client, park on this server's
own bell until the latch clears, retry. A server that does not publish
kCapRunAheadApply keeps P5C's Fatal byte for byte.
- Two named refusals survive, and they are the impossible cases rather than the busy
ones: a single event larger than half the ring (no drain can ever make room) and a
30 s wait for a client that is not draining because it is not running.
- The deadlock argument needs both halves, so both are here.
EventRingConsumer::Drained() now CLEARS AND RINGS - a cleared flag with no bell is
the lost wakeup the forward direction's publish-then-ring order exists to prevent -
and SessionProducer gains WaitForAppliedOrEventBacklog /
WaitForPresentAckOrEventBacklog so a parked client is always wakeable by the server
that is waiting for it. The old two waits keep their callers: teardown has no ring
to drain and must not learn about one.
- ServerLoop's park condition gains a fourth clause, and it is the one that says "not
now" rather than "wake up": with the ring full a waiting record is not work yet,
because applying it would only park the producer half way through. Stop and Control
still win, so a teardown is never held up by a client that stopped draining.
- Inert this phase: nothing publishes kCapRunAheadApply until the integration commit,
so every arm above answers as it did before.
|
||
|
|
1f8de61be1 |
[MG_Remote, MG_Pipe, MG_Impl, MG_State, MG_Backend, MG_Test, MG_IntegrationTest] (Disaggregated): P5d round 3 - the profile-directed cuts on both roles
- The round-2 report guessed the apply thread's ~48% mutex time was the backend
pendingMutex; the symbolized call graph at
|
||
|
|
cb06538c06 |
[MG_Remote, MG_Pipe, MG_Impl, MG_Test, MG_IntegrationTest, Docs] (Disaggregated): inproc performance package - dirty-driven persistent-map push, batched barrier waits, spin-clock batching
VD12 Minecraft on the Redmi (FCL, DirectGLES): the split's in-world frame rate
13 -> 60-66 fps against monolith's 110-114, and persistent-map wire bytes
25 MB/frame -> ~0.23 MB/frame. The residual split-specific costs are named at
the end, not hidden: the R-1 serialization (WaitForApplied ~12%, P11's
versioned-inputs design) and a split-only mutex complex still under
investigation.
Persistent-map push, three arms picked per buffer at map time:
- mprotect dirty tracker: the mapped range's page-aligned interior is
protected PROT_READ; the app's next write faults, the SIGSEGV handler marks
the page in the slot's bitmap and unprotects, and the push ships one 64 KB
block per block containing a faulted page. Ownership is three tests -
SEGV_ACCERR only (a userspace raise and an unmapped page chain verbatim),
inside a live tracked range, and not a refault on a still-set bit (a write
we unprotected succeeds, so that is a non-write access and it crashes
honestly). The alignment is INWARD: an outward-aligned edge page holds
foreign bytes, and a signal-blocking worker thread faulting on one is a
process kill - that was the ctest-only SmallRing segfault. Edge spans
(at most two per map) are hashed per push and ship their containing block
only on change (unconditional edges measured 112 MB/frame). Re-arm is one
mprotect per contiguous run. The first push ships everything once.
- hash-suppress arm (MOBILEGL_IPC_PERSISTENT_HASH_SUPPRESS, default 1):
per-block XXH3 against the last push's hashes; untracked buffers only.
- PushDrawConsumers: only the buffers the upcoming draw/dispatch can read
(VAO attribs/element, indexed binding points, indirect/parameter slots,
TexBuffer backing - the last also retires the per-frame-cadence cloud
twitch), plus the fault-epoch skip: an unmoved epoch proves no interior
byte changed, tracked edges are hashed in place, and any hash-arm member
vetoes. The membership map caches the tracker slot pointer (a 64-slot
rescan per buffer per draw was measured and removed).
MOBILEGL_IPC_BATCH_WAITS (default 1; MOBILEGL_PIPE_VERIFY forces 0): a
value-class record (kCtxState/kCtxCso/kCtxObject with no reply slot) is
published without waiting for its own apply; the barrier is taken at the next
pull-reading verb, the only place BARRIER-PULLED fields are read, so the R-1
fence posts are untouched and only the round trips between them change. Two
consequences are pinned rather than argued: kCtxObject verbs the tests read
synchronously get explicit fences (GenerateMipmapFieldsCross,
FramebufferDeathCrosses), and a frontend object whose last SharedPtr drops on
the apply thread - impossible before batching - now enqueues its
(kind, lifetimeId) on the deferred destroy queue and the GL thread replays
the DestroyAndFree helper at the verb hooks (CopyImageLayeredScenario aborted
Fatal{RoleViolation, MGPipeSlots} without it).
Doorbell::Wait's spin reads the clock once per 64 yields: on the workload
device a steady_clock read is a real syscall (no vDSO) and the per-iteration
reads were ~20% of the whole process at ~1000 waits/frame.
Gates: gen_pipe --check/--self-test green; unit 2188/2188;
integration-split 110/111, the one red being GenerateMipmapDepthPixels whose
DirectGLES and DirectVulkan MONOLITH siblings flake identically with every
code path of this package provably inert (PushIsArmed false, no ClientSession,
deferral's monolith early-out).
|
||
|
|
930ad0193f |
[MG_Remote, MG_Impl, MG_Backend] (Disaggregated): P5c ev - SEG_EVENT is the only reverse channel
The three reverse MGPipeCallbacks entries become the SERVER session's producer callbacks
(Reserve + fill the head + copy the payload inline + PublishEvents; the host pointer never
reaches the wire), installed at Accept and uninstalled at Close, a second installation
over a claimed entry being Fatal{RoleViolation, callback-double-install}. Producers wired:
buffer writeback (the mapped bytes are copied into the record's inline tail instead of a
raw pointer in MGPBlobRef::Offset), GPU-written (Magma's three direct MarkGpuWritten
bypasses are routed through the callback), surface-changed (the backend posts
MGPSurfaceInfo; the client consumer replays the allocate/format writes against
pDefaultFramebufferInfo on the GL thread, where R3's ownership always was), and
kEventGlError = 4 for PipeInputs::RecordError (ordering stays P9's).
PipeInputs::InvalidateCompileEnv's forward is deleted with an active transport - R-12's
caps re-publication already does its job. The consumer's monolith guard is replaced by
the three segment arms (kSegEvent resolves through the session's SegmentTable;
kMGHostSpanSegNone is monolith-only and Fatal{ProtocolCorruption} on the wire; anything
else the same), and DrainEventRing calls the client consumers BY NAME, never through the
global table. Overflow is Fatal{EventRingOverflow} - P5c's events are lossless, P9 owns
the drop policy. Monolith arms keep the pre-P5c expressions character for character.
Evidence: unit 2139/2139 (incl. the new kEventGlError round-trip), integration-split
107/107, monolith GPU subset 24/24; negative control executed - a writeback blobref
reverted to the raw-pointer shape turns DirectGLES.Split.AtomicCounterScenario.
SubDataAfterDispatchSurvivesAnImmediateReadback red with Fatal{ProtocolCorruption,
OnBufferWriteback.Seg}. CONTRACT-P5C sections 1, 4.
|
||
|
|
949ba509f3 | [Fix, Test] (MG_Remote, MG_Test, CONTRACT): wave 1.5 s1 - ID-47 SEG_REPLY is 16 MiB / 8 slots of 2 MiB and ReplySlotPool::CanHold+RequireReadPixelsFits refuse an oversize ReadPixels at the CLIENT by name (ProtocolSmokeTest pin, CONTRACT row 23); CapsAbiFingerprint IS MixAbiFingerprint over CapsAbiFingerprintInputs and the sensitivity case drives it; the null-union Hello/Welcome frames go THROUGH Accept/StartOverTransportPair; the pad-bit case stamps the bit past Reserve; both death controls name their diagnostic via WireLogFatal (stderr echo); the RSS ledger reads /proc once and keeps its own running peak (GitHub run 35079459114) | ||
|
|
f33e5d6090 | [Fix] (MG_Remote, Transport, Client): SEG_STAGE is w1's encoder-local linear allocator and not a ring - drop the stage RingProducer and its power-of-two rounding, pin the stage cursor triple dead, require a wrap filler to carry kind kRingPadRecordKind as well as the flag it shares with kVarTail, and name every kRecBorrowSlot sighting | ||
|
|
da8f030459 | [Fix] (MG_Remote, Transport): the knob names the ring and the segment adds a control page, a backwards watermark is Fatal rather than a permanent hang, RetireThrough stops at a borrowed slot, the inproc peer gets a real second mapping through dup and Adopt, and the producer remembers its own published seq | ||
|
|
817e1c40cf | [Feat] (MG_Remote, Transport): the ring-owning session pair over ShmSegment - four segments, the five watermarks with their first real writers, the SEG_REPLY slot pool and the SEG_EVENT reverse channel, with the doorbell pair taken from the transport and the ring capacity derived as the largest power of two left after the control page | ||
|
|
245ad14e6c | [Docs, Test] (MG_Remote, Wire): state the five sequence watermarks, the late-never-early batching rule and the pad-does-not-count rule on RingControl, and pin all three with five cases | ||
|
|
901d48a678 |
[Fix] (MGPipe, Metrics, Config, CI): exchange the per-frame stats instead of racing a store, carry the three uncarried table entries, drop the inline host span from a buffer range, spell the buffer subdata range, and close the small gate holes
- S-1 PipeStats::OnPresent read each frame accumulator and then store(0)'d it; a Bump from a staging thread landing in between was lost from the Tracy plot and from every frame. Each accumulator is now exchange(0, relaxed) and the exchanged value is what is plotted, so every add lands in exactly one frame. - T-3 FdPassing without MSG_CMSG_CLOEXEC (macOS, BSD) handed back descriptors that survived exec; every received fd now gets FD_CLOEXEC by hand under !MSG_CMSG_CLOEXEC. MSG_NOSIGNAL is defined to 0 where the platform lacks it (FdPassing.cpp, Doorbell.cpp) and SO_NOSIGPIPE is set on the socketpair and on a SocketDoorbell's descriptor where it exists, so a write to a hung-up peer is EPIPE rather than a fatal signal. - T-4 the missing-flatbuffers fallback wrote OFF into the cache with FORCE, so a plain re-configure after `git submodule update` stayed OFF silently. It is a normal-variable set now, shadowing the cache for that configure only; verified by hiding flatbuffers.h, configuring with ON (warning, transport off, cache still ON) and re-configuring plainly with the header back (transport ON). - P-2 three LIVE GLFunctionsTable entries had no carrier: GetGpuTimestampNs (glGetInteger64v(GL_TIMESTAMP), a synchronous server answer), QueryCounterTimestamp (glQueryCounter, a one-shot stamp, not a begin/end pair) and WaitSync (the GPU-side wait FenceWait's client wait does not express). QueryTimestamp (MGPTimestampRequest, kCtxQuery, kReplySlot), QueryCounter (MGPQueryDesc with Kind = GL_TIMESTAMP, kCtxQuery) and FenceWaitServer (MGPFenceWait, kScreen) are APPENDED at the end of PipeCalls.def because the opcode is the position: SetSwapInterval stays 68, the three take 69-71, and PipeCatalogue.LateArrivalsAreAppendedWithoutRenumbering pins that. Header counts 71 (screen 11, query 8); the seven generators regenerated. - P-3 MGPBufferRange inlined a 32-byte MGHostSpan into every range of every class - dead space on every SSBO, atomic-counter and XFB range, and D-B8 says not to freeze the named-UBO payload before the stage-ubo-named numbers exist. The range is 24 bytes now; the host spans are an optional second var-tail behind the ranges, announced by MGPShaderBuffers::HostSpanCount (0 or Count), with set_shader_buffers keeping its kVarTail|kHostSpan flags. PipeCatalogue.BufferRangeCarriesNoInlineHostSpan pins the sizes, the flags and the comparator's view of the count. - P-4 QueryEnvUint64 parsed with base 0 (a leading zero meant octal: MOBILEGL_PIPE_PUSH=010 read as 8) and accepted -1 as every bit set; it is decimal or explicit 0x now and a '-' anywhere is refused with the warning (smoke through the integration binary: -1 and 12abc warn, 010 and 0x10 parse). The CI stdio gate's alternation now also catches fprintf(stdout, puts( and std::cout/cerr; it is green over MG_Backend and MG_State. MGPSubData states how the buffer half expresses [offset, size): UnionBox.X / UnionBox.W with Target == Buffer, Y = Z = 0, H = D = 1, one record bounded at a 2^31-1 offset and 2^32-1 size beyond which the emitter splits (the same rule the ring's half-capacity bound already imposes); MGPipeSetSubDataBufferRange / MGPipeSubDataBufferOffset / Size are the only spelling and PipeCatalogue.SubDataBufferRangeRidesInTheUnionBox pins the encoding and its bounds. gen_pipe.py now refuses, in both modes, a call payload named in PipeCalls.def with no field list in PipeFields.def (the four memcmp-fallback member types are the documented exception); shown by dropping P(MGPSwapInterval), which exits 1 naming the payload. - The MGPPixelPackState size assertion compared sizeof against itself; it asserts the literal 28 PixelStoreParameters measures. - Verified: ctest -L unit green in both the default and the split configuration, gen_pipe.py --check clean with the generated files committed, nm --defined-only of the default libMobileGL.so has no MG_Remote symbol, and the full integration-gpu suite passes (the *IsActuallyArmedWhenTheEnvironmentPinsItOn family trips under -j 8 as documented and passes serially). |
||
|
|
1154f9a00d |
[Fix] (MG_Remote, Transport): give the inproc doorbell a death state so Shutdown can join a parked waiter, and bound a ring record at half the capacity so a refusal can never look like backpressure
- T-1: InProcessChannel::Close rang each CondVarDoorbell once and claimed that unparks a peer mid-frame. It does not. Doorbell::Wait consumes the one ring, re-tests a condition nothing published, finds the bell alive (CondVarDoorbell never overrode Dead(); Doorbell.cpp had no death state at all) and with kWaitForever parks again for good - so Shutdown could never join a server thread sitting in the design's own steady state (spun, set consumerParked, blocked; plan section 8.1 inheriting the earlier plan's 6.2a). CondVarDoorbell now carries an atomic death latch: Kill() sets it under the mutex and notify_all's, Dead() reports it, Park returns false at once on a dead bell (and the wait predicate includes it, so a Kill cannot slip between the test and the wait), and Close kills both bells instead of ringing them. Same shape as SocketDoorbell's EOF latch; Notify stays the ordinary wakeup. - T-2: RingProducer::Reserve refused only total > capacity, but a record with capacity/2 < total <= capacity is unplaceable at every head offset where neither the space to the wrap boundary nor the space before it holds it - even in an EMPTY ring, because a wrap pad costs spaceToEnd bytes on top of the record. Concretely: head offset 16 of an empty 256-byte ring, a 248-byte record; FreeBytes() says 256, Reserve says nullptr, forever, and a producer waiting for FreeBytes() >= 248 stalls with nothing logged. The bound is now capacity/2, which is exact rather than conservative (worst case 2*total-8 <= capacity-8), exposed as MaxRecordBytes() for the emitter to chunk against; the minimum ring is two headers so the smallest record still fits the bound. Ring.h states Capacity()/2 as the chunking bound and the G3 header comment in gen_pipe.py now states the chunking rule plan section 8.2 asks G3 to define (PipeWire.inc regenerated). - Tests, each shown red with only the fix site reverted and green with it: InProcessTransportTest.ShutdownUnparksAWaiterWithNoDeadline (bounded join through a shared_ptr-owned waiter: 5 s red instead of a hung job; reverted it hangs and fails at 5051 ms), RingTest.RecordLargerThanHalfTheRingIsRefused (reverted, the 248-byte record is accepted), RingTest.RecordPlaceabilityDoesNotDependOnTheHeadOffset (the offset-0 vs offset-16 negative control), RingTest.HalfCapacityRecordFitsAtEveryHeadOffset (the positive half: the maximal record at all 32 head offsets of a 256-byte ring) and RingTest.RejectsARingTooSmallForTheSmallestRecord. |
||
|
|
c1a7ffac94 |
[Fix] (MG_Remote, Transport): give descriptor offers their own condition variable, cap the ring at what a 32-bit size field can describe, and keep the frontend umbrella out of Framing.h
- InProcessChannel::Direction served two different predicates from one
condition_variable signalled with notify_one, so a SendFrame wakeup could be
delivered to a thread blocked in ReceiveFd, which re-tests its own predicate
and goes back to sleep - leaving a queued message undelivered until some
unrelated later event. ITransport narrows the contract to one dedicated
reader thread, but that is a comment, not a mechanism, and the first caller
that splits its reader should not have to discover this. Offers now have
their own fdCv, and Close() notifies both.
- RingProducer/RingConsumer accepted any power-of-two capacity while
RingRecordHeader::size is 32-bit by wire contract (plan section 8.1 ->
PLAN.md section 6.3: 8-byte RecHeader). At 4 GiB or more a record's size -
or a wrap filler's, which is sized by the distance to the boundary - would be
truncated on the way in, and the consumer would then bounds-check the
truncated value against the real one. kMaxRingCapacity rejects that at
construction, the same class of guard as the power-of-two and
smaller-than-a-header checks beside it. Unreachable today (SEG_CMD 8 MiB,
SEG_STAGE 32 MiB), which is the point of catching it now.
- Framing.h included MG_Util/Debug/Log.h, which includes Includes.h, the GL
frontend's umbrella header: 661 headers by `clang++ -H`. It is the one header
under Transport/ that broke the rule ITransport.h states for this layer
("nothing about a byte pipe needs the GL frontend's umbrella header"), which
matters when the server-side binary links this and when the include-graph
purity gate of plan section 10.3 (gate A, asserted on -H output rather than
on symbols) lands. Its three error paths now call WireLogError, declared in a
new dependency-free WireLog.h whose .cpp owns the umbrella. Framing.h is down
to 134 headers, none of them MG_State, Includes.h or Log.h.
- Two documentation corrections. ITransport::Shutdown documented a one-sided
"releases the endpoint" while InProcessTransport::Shutdown closes both
directions - which is what closing a socket does, so the spawn transport will
behave the same way; the interface now says whole-connection teardown, and
keeps the promise that queued messages stay readable until drained.
InProcessTransport.h cited a CMake option
MOBILEGL_BUILD_DISAGGREGATED_INPROC that grep finds nowhere: plan appendix B
reserves it for the role-isolation shim, this skeleton does not add it, and
the delivery mode is a runtime choice (MOBILEGL_TRANSPORT), not a build one.
- Evidence: both configurations reconfigured and rebuilt; nm --defined-only on
the OFF build still reports 0 MG_Remote symbols and the ldd dependency set is
byte-identical to the OFF link (plan section 10.3, the two surviving
byte-level equalities). Negative controls: removing the capacity ceiling
makes RingTest.RejectsACapacityTheRecordHeaderCannotDescribe fail on both
roles; collapsing fdCv back into cv makes
InProcessTransportTest.AFrameWakeupIsNotEatenByAWaiterOnDescriptors fail at
3950ms against its 2000ms bound.
|
||
|
|
bdd4bed431 |
[Fix] (MG_Remote, Transport): close the doorbell's lost-wakeup window with two seq_cst fences and stop a hung-up peer turning every park into a spin
- The header claimed the park flag's own seq_cst store and load closed the lost-wakeup window. They cannot: that Dekker argument needs all FOUR accesses in the seq_cst total order, and the other two are not in it - the watermark publish is a release store (RingProducer::Publish) and the condition re-test is an acquire load. There was no atomic_thread_fence anywhere under MG_Remote. On x86 a release store and a seq_cst load are both plain MOVs, so the notifier can read parked==0 while its head store still sits in the store buffer, and the waiter then parks on a stale watermark forever; ARMv8 survived only because STLR->LDAR is RCsc, which is luck, not the design. Doorbell::Wait now fences after announcing and NotifyIfParked fences before reading the flag - the pairing the standard actually guarantees ([atomics.order]) - and the header says so, including the other half of the contract: publish the watermark BEFORE ringing, because a fence only orders what precedes it. This is the claim the whole P5/P6 wait discipline (present credit, kNeedsAck blocking requests, full-ring escalation) will be built on, and its failure mode is a silent cross-process hang. Inherited design, plan section 8.1 (PLAN.md section 6.2/6.2a: bidirectional doorbell, MOBILEGL_IPC_SPIN_US default 50us, condvar for inproc). - SocketDoorbell::Park treated any poll() return > 0 as a wakeup and never looked at revents. Measured on this machine: an AF_UNIX SOCK_STREAM socketpair whose peer has closed returns revents=POLLIN|POLLHUP with recv()==0 immediately and forever. Park therefore returned true, Wait stored parked=0, found its condition still false and re-parked - so a waiter with kWaitForever burned a big core at full clock with no bound. That is exactly the pathology the bidirectional doorbell exists to prevent (a whole 16.6ms frame of a big core on a phone, competing with the GPU and the game's JVM), reached from the other side. Park now branches on revents, a new Drain() latches death on EOF (and on ECONNRESET/EPIPE from Notify), and the new Doorbell::Dead() lets Wait give up instead of re-parking on a descriptor that can never deliver another wakeup. - Same commit, same defect class: `fd` is documented as one end of an AF_UNIX socket pair, not "a socket or pipe end". Notify uses send(MSG_DONTWAIT| MSG_NOSIGNAL) and Park uses poll()+recv(), which a pipe end refuses with ENOTSOCK, and a SOCK_DGRAM pair reports no readiness at all when the peer closes (measured), so the spawn transport wants SOCK_STREAM. - Evidence: build-linux-split rebuilt clean; the wire suite is green (47/47). Negative control - reinstating the revents-blind Park makes FdPassingTest.SocketDoorbellStopsParkingWhenThePeerHangsUp fail on both Park assertions, and restoring this code turns it green again. |
||
|
|
bfa087d0f7 |
[Feat] (MG_Remote, Transport): land the transport skeleton behind MOBILEGL_BUILD_DISAGGREGATED - framing, the SPSC ring, doorbells, shm segments and SCM_RIGHTS
- New CMake option MOBILEGL_BUILD_DISAGGREGATED (default OFF, plan B appendix B). ON appends the MG_Remote sources to SOURCE_FILES, puts 3rdparty/flatbuffers/include on the include path and defines MOBILEGL_BUILD_DISAGGREGATED=1. OFF compiles nothing from MG_Remote and adds no include path and no library, which is one of the two byte-level equalities plan B section 10.3 keeps: measured `nm --defined-only build-linux/libMobileGL.so | grep -ic MG_Remote` = 0 with the option OFF and 93 with it ON, with an identical ldd set in both configurations.
- The option forces itself OFF with message(WARNING) when 3rdparty/flatbuffers/include/flatbuffers/flatbuffers.h is missing, so a checkout without the submodule still configures and builds rather than failing a hundred lines later on a missing header.
- ITransport.h: SendFrame / ReceiveFrame / PeekFrameSize / ShareFd / ReceiveFd / Shutdown. ReceiveFrame's contract is the fix for a defect of the earlier branch: a destination buffer smaller than the pending message returns MOBILEGL_ERR_BUFFER_TOO_SMALL with the required size and KEEPS the message queued. The earlier transport failed the call and popped the message anyway, which wedges the stream permanently the first time a reader guesses a size wrong.
- Framing.h: [u32 'MGLF'][u32 len][payload], 64 MiB cap, validated on read. A bad magic or an oversized length latches the reader dead, is logged at ERROR and makes every later call return MOBILEGL_ERR_PROTOCOL_MISMATCH. This is the second inherited defect: Feat/CS-Delta-IPC's Framing.h:41-45 Feed() always returned OK and its header peek merely returned false, so a desynchronized stream became a silent permanent hang; its LocalSocketTransport.cpp:232-236 then allocated on the peer-supplied length with no cap. The magic is also byte-ordered so it reads "MGLF" on the wire, where the earlier constant spelled "FLGM".
- Ring.h/.cpp: RingControl exactly as the inherited design (plan B section 8.1 -> earlier section 6.2): two independent cursor triples (cmd and stage, each {head, appliedTail, retiredTail}), appliedSeq / submittedSeq / retiredSeq / completedFrameSerial / presentAckSerial, serverEpoch, ringGeneration, consumerParked, producerParked, eventRingFull, eventDropped. One 4096-byte page with each contended group on its own cache line, pinned by static_assert on size and alignment and by an offset test.
- The SPSC pair uses monotonic byte cursors and a power-of-two mask, so a torn read can never look like a valid earlier position. A record never straddles the wrap: the producer emits a kPad filler to the boundary, which is always a multiple of 8 and therefore always has room for a header. The producer reclaims against retiredTail rather than appliedTail, so the day the server borrows a ring slot into the GPU timeline (kRecBorrowSlot) it does not silently degrade to early recycling. HardDrainRing bumps ringGeneration only when the ring is quiesced, and leaves the cursors monotonic so cached offsets are recognisably stale.
- The consumer bounds-checks every record header before dispatch (8-aligned, at least a header, no larger than what the producer published, contiguous inside the mapping) and reports corruption to the caller instead of dispatching into undefined behaviour. That is the runtime half of the earlier section 6.3 discipline: SEG_CMD is written by another process, so a compile-time static_assert on record sizes proves nothing about what is in the mapping.
- Doorbell.h/.cpp: spin then park, both directions (earlier section 6.2a). CondVarDoorbell for inproc, SocketDoorbell for spawn (one byte, codes 0x01 ring-advanced and 0x02 watermark-advanced) - no futex, eventfd or named event anywhere. The lost-wakeup window is closed by ordering: the waiter stores its park flag seq_cst and then re-tests the condition, NotifyIfParked loads the same flag seq_cst after the watermark is published, so one of the two always sees the other. kDefaultSpinUs is 50, the MOBILEGL_IPC_SPIN_US default.
- ShmSegment.{h,cpp} + ShmSegmentPosix.cpp: memfd_create by raw syscall on desktop Linux (the glibc wrapper is too recent to rely on), ASharedMemory_create on Android (API 26; libc's memfd_create wrapper is API 30, above MobileGL's floor), shm_open + immediate shm_unlink as the fallback. Adopt() refuses a descriptor whose fstat size is smaller than the size the peer announced, so a short segment cannot turn every later offset into an out-of-bounds map. ShmSegmentWin32.cpp (CreateFileMappingW in Local\) is compile-guarded and untested - this project's Windows machine is not a correctness gate.
- The Android path is compile-verified, not just written: an arm64-v8a NDK build with the option ON links, ShmSegmentPosix.cpp.o carries an undefined ASharedMemory_create, and ShmSegmentWin32.cpp.o is empty there. That build is also what caught the missing <cstddef> in ShmSegment.h and Framing.h, where std::size_t / std::ptrdiff_t only resolved through a transitive include on the host sysroot.
- FdPassing.{h,cpp}: SCM_RIGHTS in the FIRST transport commit, as plan B section 8.1 demands, over a dedicated AF_UNIX SOCK_DGRAM socketpair rather than the control byte stream (message boundaries survive on every POSIX - SOCK_SEQPACKET does not exist on macOS - and ancillary data can never be split from its payload). The third inherited defect this replaces: Feat/CS-Delta-IPC deferred fd passing and hardcoded `out->fd = -1` in LocalSocketTransport.cpp:296, so on the only platform that matters its data plane could not move one byte between processes. MSG_CTRUNC, an unexpected descriptor count and a malformed sideband header all close every descriptor received before failing, and a too-small sideband buffer is refused before the recvmsg so a datagram is never half-consumed.
- InProcessTransport.{h,cpp}: two in-memory queues plus the condvar doorbell pair. It keeps the frame size cap so nothing that passes in inproc becomes illegal after the switch to spawn, hands descriptors over with dup() under the same ownership rule as SCM_RIGHTS, and lets a peer's queued messages be drained after Shutdown - usually the last one says why it is going away.
- Not done here on purpose: the MOBILEGL_IPC_* environment variables are parsed in ConfigLoader.cpp, which belongs to another P0 work package running in parallel; this commit only exposes the constants (kDefaultSpinUs, kMaxFramePayloadSize, FdPassing::kMaxSidebandBytes) so that plumbing has something to set.
|