paulmckrcu / litmus Goto Github PK
View Code? Open in Web Editor NEWAutomatically generated litmus tests for validation LISA-language Linux-kernel memory models
License: GNU General Public License v2.0
Automatically generated litmus tests for validation LISA-language Linux-kernel memory models
License: GNU General Public License v2.0
I'm trying to run some of the litmus tests from this repo using herd7
(version 7.56
) with the cat models (and corresponding bell files) also in this repo, but I'm getting problems due to unbound vars, e.g.
herd7 auto/LB-GWR+R-Dd+R-Dd+R-Oc+OB-OB.litmus -model models/alan-kernel-2016.04.28.cat -bell models/alan-kernel-2016.04.28.bell
File "./models/alan-kernel-2016.04.28.bell", line 29, characters 27-32: unbound var: Start
The other models have the same problem (always Start
). I took a quick look to the herd7
source code (main
branch), but could not find anything related to this kind of "events".
Any idea what the problem might be? Am I missing any flag or do I need some special version of herd7
?
The following program has a potential related-acq pair (P0 -> P2
) that can be broken by an intervening thread (P1
) with a relaxed instruction
❯ cat weird-C.litmus
C weird
{}
P0(intptr_t *x, intptr_t *y)
{
WRITE_ONCE(*x, 1);
atomic_set_release(y, 1);
}
P1(intptr_t *x, intptr_t *y)
{
int r0 = atomic_fetch_add_relaxed(2, y);
}
P2(intptr_t *x, intptr_t *y)
{
int r0 = atomic_read_acquire(y);
int r1 = READ_ONCE(*x);
}
exists (2:r0=3 /\ 2:r1=0)
The LKMM from here allows the chain to be broken
❯ herd7 -model ../lkmm/tools/memory-model/linux-kernel.cat -macros ../lkmm/tools/memory-model/linux-kernel.def -bell ../lkmm/tools/memory-model/linux-kernel.bell weird-C.litmus -I ../lkmm/tools/memory-model/
Test weird Allowed
States 7
2:r0=0; 2:r1=0;
2:r0=0; 2:r1=1;
2:r0=1; 2:r1=1;
2:r0=2; 2:r1=0;
2:r0=2; 2:r1=1;
2:r0=3; 2:r1=0;
2:r0=3; 2:r1=1;
Ok
Witnesses
Positive: 1 Negative: 9
Condition exists (2:r0=3 /\ 2:r1=0)
Observation weird Sometimes 1 9
Time weird 0.01
Hash=f3a36a6d5313d0bcee071c2a06783c41
The chain could be fixed by making the RMW in P1
a release
❯ cat weird-C-fixed.litmus
C weird
{}
P0(intptr_t *x, intptr_t *y)
{
WRITE_ONCE(*x, 1);
atomic_set_release(y, 1);
}
P1(intptr_t *x, intptr_t *y)
{
int r0 = atomic_fetch_add_release(2, y);
}
P2(intptr_t *x, intptr_t *y)
{
int r0 = atomic_read_acquire(y);
int r1 = READ_ONCE(*x);
}
exists (2:r0=3 /\ 2:r1=0)
❯ herd7 -model ../lkmm/tools/memory-model/linux-kernel.cat -macros ../lkmm/tools/memory-model/linux-kernel.def -bell ../lkmm/tools/memory-model/linux-kernel.bell weird-C-fixed.litmus -I ../lkmm/tools/memory-model/
Test weird Allowed
States 6
2:r0=0; 2:r1=0;
2:r0=0; 2:r1=1;
2:r0=1; 2:r1=1;
2:r0=2; 2:r1=0;
2:r0=2; 2:r1=1;
2:r0=3; 2:r1=1;
No
Witnesses
Positive: 0 Negative: 9
Condition exists (2:r0=3 /\ 2:r1=0)
Observation weird Never 0 9
Time weird 0.01
Hash=2d1ad2a445f148efe7634b8992363e5d
This behaviour would neither be observed in an ARM
❯ cat weird-ARM.litmus
AArch64 weird
{
0:X1=x; 0:X3=y;
1:X1=x; 1:X3=y;
2:X1=x; 2:X3=y;
}
P0 | P1 | P2;
MOV W0,#1 | MOV W0,#2 | LDAR W0, [X3];
STR W0,[X1] | LDXR W2,[X3] | LDR W2, [X1];
STLR W0,[X3] | ADD W2,W2,W0 |;
| STXR W1,W2,[X3] |;
exists
(2:X0=3 /\ 2:X2=0)
❯ herd7 -model cat/aarch64.cat weird-ARM.litmus
Test weird Allowed
States 6
2:X0=0; 2:X2=0;
2:X0=0; 2:X2=1;
2:X0=1; 2:X2=1;
2:X0=2; 2:X2=0;
2:X0=2; 2:X2=1;
2:X0=3; 2:X2=1;
No
Witnesses
Positive: 0 Negative: 15
Condition exists (2:X0=3 /\ 2:X2=0)
Observation weird Never 0 15
Time weird 0.01
Hash=342601047a6d69dee710e4e94d28e4fe
or PowerPC machine
❯ cat weird-PPC.litmus
PPC weird
{
0:r2=x; 0:r4=y;
1:r2=x; 1:r4=y;
2:r2=x; 2:r4=y;
}
P0 | P1 | P2 ;
li r1,1 | lwz r3,0(r4) | lwz r1,0(r4) ;
stw r1,0(r2) | cmpw r3,r3 | cmpw r3,r3 ;
stw r1,0(r4) | beq LC00 | beq LC01 ;
| LC00: | LC01: ;
| addi r3,r3,1 | isync ;
| stw r3,0(r4) | lwz r3, 0(r2) ;
exists
(2:r1=3 /\ 2:r3=0)
❯ herd7 -model cat/power.cat weird-PPC.litmus
Test weird Allowed
States 6
2:r1=0; 2:r3=0;
2:r1=0; 2:r3=1;
2:r1=1; 2:r3=0;
2:r1=1; 2:r3=1;
2:r1=2; 2:r3=0;
2:r1=2; 2:r3=1;
No
Witnesses
Positive: 0 Negative: 18
Condition exists (2:r1=3 /\ 2:r3=0)
Observation weird Never 0 18
Time weird 0.01
Hash=09d4d22cf368195d3c18a8c9acf34a2c
My question is then why the LKMM allows this behaviour? Is there any weaker architecture supported by the kernel (e.g. Alpha) where this behaviour would be observable?
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.