Refuse a streamed file that commits more than it reserved
osFileStart sets an extent aside and osFileWrite refuses a block index outside it, so writing off the end was already barred. Committing a larger size was not, and reaches the same neighbour by simply claiming it: a directory entry is the only record of what a file owns, so an entry claiming a block it was never given owns it, and so does whatever owned it before. Both files then look perfectly well formed. The free count went backwards past zero on the same path. Found by ChatGPT's review of the streaming work, in NOTES.md. I had bounded the index because writing off the end was the obvious way to reach a neighbour, and had not noticed that the other end of the same reservation was open. THE SIZE IS COMPARED, NOT THE ROOM IT TAKES UP. One block and a tail occupies exactly what two whole blocks occupy, so bounding the blocks alone would let a file reserve the first, commit the second, claim no block it was not given, and still report two hundred and forty six bytes that were never written to it - whatever the disk had there before. Checked before anything is touched, which is why the temporary is found twice. The old file is deleted a few lines down and a refusal after that point would have destroyed the thing it was protecting. AND IT CAUGHT A REAL ONE IMMEDIATELY. The assembler reserves the file plus room for its vectors, and asked for four bytes per vector DECLARED - which looks like a safe bound and is not, because a device is declared during the SECOND pass, in the line that implements it. A program with a device installs a vector that was not counted when the room was measured. CosmOS reserved 14,163 bytes and committed 14,167, writing four bytes past what it had been given on every build since S2. It landed inside the last block it owned, and would not have if the boundary had fallen four bytes earlier. It reserves against the vector table's LIMIT now, which cannot go stale whenever things are counted. Claim.asm is the program that tries it: reserve one block and a tail of ten, write them, then tell osFileDone the file came to two whole blocks. The refusal and the honest commit that follows are both recorded. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01E2JrLzFvuFX9fgi1LDRjrW
This commit is contained in:
Anachronaut
co-authored by
Claude Opus 5
Claude Opus 5
parent
19ab36a201
commit
a7d3e09d94
@@ -0,0 +1,102 @@
|
||||
; Tries to commit a file bigger than the room it reserved.
|
||||
;
|
||||
; osFileStart sets aside an extent and osFileWrite refuses a block index outside it, so the
|
||||
; obvious way to reach a neighbouring file - writing off the end - is already barred. This
|
||||
; is the other way to the same place: reserve one block, write the one block, and then tell
|
||||
; osFileDone the file came to two.
|
||||
;
|
||||
; NOTHING WOULD SAY SO IF THAT WERE ALLOWED. A directory entry is the only record of what a
|
||||
; file owns, so an entry claiming a block it was never given simply owns it, and whatever
|
||||
; owned it before owns it too. Both files then look perfectly well formed.
|
||||
;
|
||||
; Correct behaviour is a refusal, and the file left as it was. The reservation is one block
|
||||
; and a tail of ten, so:
|
||||
;
|
||||
; two blocks and no tail more than was reserved refused
|
||||
; one block and a tail exactly what was reserved allowed
|
||||
;
|
||||
; Written by Anachronaut
|
||||
|
||||
#Include services.asm
|
||||
#Program
|
||||
#Base 0x4000
|
||||
|
||||
start:
|
||||
; One block, and ten bytes after it.
|
||||
SETD.0 Name
|
||||
SETD.3 0x00 0x01
|
||||
INIA 0d10
|
||||
SWI osFileStart
|
||||
BNQ noStart
|
||||
|
||||
SETD.1 Block
|
||||
RSTA
|
||||
RSTB
|
||||
SWI osFileWrite
|
||||
BNQ noWrite
|
||||
SETD.1 Block
|
||||
RSTA
|
||||
INIB 0d1
|
||||
SWI osFileWrite
|
||||
BNQ noWrite
|
||||
|
||||
; Two whole blocks, which is more than one block and a tail.
|
||||
SETD.3 0x00 0x02
|
||||
RSTA
|
||||
SWI osFileDone
|
||||
BNQ refused
|
||||
SETD.0 Allowed
|
||||
SWI osPrintString
|
||||
SWI osExit
|
||||
|
||||
refused:
|
||||
SETD.0 Refused
|
||||
SWI osPrintString
|
||||
|
||||
; And the honest size, which is what was reserved.
|
||||
SETD.3 0x00 0x01
|
||||
INIA 0d10
|
||||
SWI osFileDone
|
||||
BNQ noHonest
|
||||
SETD.0 Honest
|
||||
SWI osPrintString
|
||||
SWI osExit
|
||||
|
||||
noHonest:
|
||||
SETD.0 NoHonest
|
||||
SWI osPrintString
|
||||
SWI osExit
|
||||
noStart:
|
||||
SETD.0 NoStart
|
||||
SWI osPrintString
|
||||
SWI osExit
|
||||
noWrite:
|
||||
SETD.0 NoWrite
|
||||
SWI osPrintString
|
||||
SWI osExit
|
||||
|
||||
#Data
|
||||
#Base 0x2000
|
||||
|
||||
Name:
|
||||
"claim.dat"
|
||||
Block:
|
||||
#Reserve 0d256
|
||||
Allowed:
|
||||
"claiming more than was reserved was ALLOWED
|
||||
"
|
||||
Refused:
|
||||
"claiming more than was reserved was refused
|
||||
"
|
||||
Honest:
|
||||
"and the size it really came to was taken
|
||||
"
|
||||
NoHonest:
|
||||
"the honest size was refused too
|
||||
"
|
||||
NoStart:
|
||||
"no start
|
||||
"
|
||||
NoWrite:
|
||||
"no write
|
||||
"
|
||||
@@ -1483,9 +1483,19 @@ checkImageRoom:
|
||||
;
|
||||
; MORE THAN THE FILE WILL COME TO, on purpose. How many vectors are actually installed is
|
||||
; not known until the second pass has resolved every handler, and by then the file has to
|
||||
; exist to be written into. What IS known now is how many vectors were DECLARED, and no
|
||||
; more than that can be installed - so the room asked for is the whole file plus four
|
||||
; bytes for each of them and five for a marker.
|
||||
; exist to be written into. So the room asked for is the whole file plus four bytes for
|
||||
; every vector the table can hold, and five for a marker.
|
||||
;
|
||||
; THE LIMIT RATHER THAN THE COUNT SO FAR, and that distinction cost a real bug. Asking
|
||||
; for four bytes per vector DECLARED looks like a safe bound and is not: a device is
|
||||
; declared during the SECOND pass, in the line that implements it, so a program with one
|
||||
; installs a vector that was not counted when the room was measured. CosmOS reserved
|
||||
; 14,163 bytes and committed 14,167, writing four bytes past what it had been given -
|
||||
; which happened to land inside the last block it owned, and would not have if the
|
||||
; boundary had fallen four bytes earlier.
|
||||
;
|
||||
; The limit cannot go stale that way. It is what the vector table holds, so no assembly
|
||||
; can install more, whenever they are counted.
|
||||
;
|
||||
; Asking for too much costs nothing but a moment: osFileDone is told what it really came
|
||||
; to and the difference goes back. Asking for too little would have meant writing off the
|
||||
@@ -1494,7 +1504,7 @@ checkImageRoom:
|
||||
SETD.2 ImgTotal
|
||||
CALL numSet
|
||||
SETD.0 ImgVecMost
|
||||
SETD.2 VecCount
|
||||
SETD.2 VecLimit
|
||||
CALL numSet
|
||||
SETD.0 ImgVecMost
|
||||
SETD.2 ImgVecMost
|
||||
|
||||
@@ -2474,6 +2474,61 @@ sbfsStreamDone:
|
||||
SETD.1 SbfsStreamNewTail
|
||||
STA.1
|
||||
|
||||
; ---- MORE THAN WAS RESERVED IS REFUSED ----
|
||||
;
|
||||
; A writer may finish smaller than it asked for, which is the whole point of being told
|
||||
; the size here. It may not finish BIGGER. The blocks after a file belong to whatever
|
||||
; comes next, so an entry claiming more than was set aside for it claims somebody else's
|
||||
; - and nothing anywhere would say so, because a directory entry is the only record of
|
||||
; what a file owns. The count of free blocks would go wrong in the same breath, the
|
||||
; subtraction below running backwards past zero.
|
||||
;
|
||||
; osFileWrite already refuses a block index past the end. This is the same bound from the
|
||||
; other side, and it was missing: the index was checked because writing off the end was
|
||||
; the obvious way to reach a neighbour, and committing a larger size reaches the same
|
||||
; neighbour by simply claiming it.
|
||||
;
|
||||
; CHECKED BEFORE ANYTHING IS TOUCHED, which is why the temporary is found twice. The old
|
||||
; file is deleted a few lines down, and a refusal after that point would have destroyed
|
||||
; the thing it was protecting.
|
||||
CALL sbfsStreamTemp
|
||||
CALL sbfsScanFor
|
||||
BNQ sbfsStreamNo
|
||||
|
||||
; What it was given, both halves of it.
|
||||
SETD.0 SbfsStreamResBlocks
|
||||
SETD.2 SbfsFileBlocks
|
||||
CALL sbfsSetWord
|
||||
SETD.0 SbfsFileTail
|
||||
LDA.0
|
||||
SETD.1 SbfsStreamResTail
|
||||
STA.1
|
||||
|
||||
; And the room that came to, kept for the free count at the end.
|
||||
CALL sbfsFileExtent
|
||||
SETD.0 SbfsStreamSpare
|
||||
SETD.2 SbfsWantBlocks
|
||||
CALL sbfsSetWord
|
||||
|
||||
; THE SIZE IS COMPARED, NOT THE ROOM IT TAKES UP. Those are not the same question: one
|
||||
; block and a tail occupies exactly what two whole blocks occupy, so a file reserving the
|
||||
; first and committing the second claims no block it was not given - and still reports
|
||||
; two hundred and forty six bytes more than were ever written to it, which are whatever
|
||||
; the disk had there before. Bounding the blocks alone would have called that fine.
|
||||
SETD.0 SbfsStreamResBlocks
|
||||
SETD.2 SbfsStreamNewBlocks
|
||||
CALL sbfsCompareWord
|
||||
BRC sbfsStreamNo ; More whole blocks than it was given.
|
||||
BNQ sbfsStreamFits ; Fewer, so the tail cannot matter.
|
||||
SETD.0 SbfsStreamResTail
|
||||
LDA.0
|
||||
SETD.2 SbfsStreamNewTail
|
||||
LDB.2
|
||||
CCF
|
||||
SUB
|
||||
BRC sbfsStreamNo ; The same blocks, and a longer tail.
|
||||
sbfsStreamFits:
|
||||
|
||||
; Now, and not before, the old one goes. It may not be there at all, which is what
|
||||
; writing something for the first time looks like from here.
|
||||
CALL sbfsStreamWhere
|
||||
@@ -2483,17 +2538,12 @@ sbfsStreamDone:
|
||||
sbfsStreamNoOld:
|
||||
|
||||
; And the temporary takes its name and its true size, which together are the whole of
|
||||
; what committing is.
|
||||
; what committing is. Found again, because deleting the old one read over the block it
|
||||
; lives in.
|
||||
CALL sbfsStreamTemp
|
||||
CALL sbfsScanFor
|
||||
BNQ sbfsStreamNo
|
||||
|
||||
; How much room it was given, before the entry that says so is changed.
|
||||
CALL sbfsFileExtent
|
||||
SETD.0 SbfsStreamSpare
|
||||
SETD.2 SbfsWantBlocks
|
||||
CALL sbfsSetWord
|
||||
|
||||
PSHD.3
|
||||
POPD.1
|
||||
DPUP.1 0d06
|
||||
@@ -2517,14 +2567,9 @@ sbfsStreamNoOld:
|
||||
CALL sbfsWriteBlock
|
||||
BNQ sbfsStreamNo
|
||||
|
||||
; And the difference goes back, which is what it was given less what it kept.
|
||||
SETD.0 SbfsFileBlocks
|
||||
SETD.2 SbfsStreamNewBlocks
|
||||
CALL sbfsSetWord
|
||||
SETD.0 SbfsStreamNewTail
|
||||
LDA.0
|
||||
SETD.1 SbfsFileTail
|
||||
STA.1
|
||||
; And the difference goes back, which is what it was given less what it kept. That can
|
||||
; no longer be negative: the check at the top refused the only case where it could.
|
||||
CALL sbfsStreamSize
|
||||
CALL sbfsFileExtent
|
||||
SETD.0 SbfsStreamSpare
|
||||
SETD.2 SbfsWantBlocks
|
||||
@@ -2557,6 +2602,19 @@ sbfsStreamNoOld:
|
||||
ADD
|
||||
RET
|
||||
|
||||
; The size the writer says it came to, put back where the extent arithmetic reads it from.
|
||||
; Said again rather than kept, because finding anything overwrites those two: they are where
|
||||
; a find describes whatever it last looked at.
|
||||
sbfsStreamSize:
|
||||
SETD.0 SbfsFileBlocks
|
||||
SETD.2 SbfsStreamNewBlocks
|
||||
CALL sbfsSetWord
|
||||
SETD.0 SbfsStreamNewTail
|
||||
LDA.0
|
||||
SETD.1 SbfsFileTail
|
||||
STA.1
|
||||
RET
|
||||
|
||||
; The directory the file is going in, and the name it will end up under. Said again before
|
||||
; each step, because every step goes to the disk and leaves the walk somewhere else.
|
||||
sbfsStreamWhere:
|
||||
@@ -2880,6 +2938,10 @@ SbfsStreamNewTail:
|
||||
0x00
|
||||
SbfsStreamSpare:
|
||||
0x00 0x00
|
||||
SbfsStreamResBlocks:
|
||||
0x00 0x00
|
||||
SbfsStreamResTail:
|
||||
0x00
|
||||
SbfsStreamAt:
|
||||
0x00 0x00
|
||||
SbfsStreamParent:
|
||||
|
||||
Reference in New Issue
Block a user