On Tue, Sep 08, 2015 at 11:17:58PM +0200, Julia Lawall wrote:
> On Tue, 8 Sep 2015, Jörn Engel wrote:
> 
> > Tried to use coccinelle on the linux kernel for the first time.  It
> > seems to get things mostly right, but not quite.  Here is the semantic
> > patch:
> > 
> > @@ expression E; @@
> > -down_read(&E->mmap_sem)
> > +mm_read_lock(E)
> > @@ expression E; @@
> > -down_read_trylock(&E->mmap_sem)
> > +mm_read_trylock(E)
> > @@ expression E; @@
> > -up_read(&E->mmap_sem)
> > +mm_read_unlock(E)
> > @@ expression E; @@
> > -down_write(&E->mmap_sem)
> > +mm_write_lock(E)
> > @@ expression E; @@
> > -down_write_trylock(&E->mmap_sem)
> > +mm_write_trylock(E)
> > @@ expression E; @@
> > -up_write(&E->mmap_sem)
> > +mm_write_unlock(E)
> 
> So this all went OK?

Mostly.  There are still a few troublemakers remaining.  But few enough
that I don't mind handling them manually.

Automatic changes:
 140 files changed, 581 insertions(+), 581 deletions(-)

Troublemakers:
arch/mips/mm/fault.c (5 lines)
arch/um/include/asm/mmu_context.h (2 lines)
include/linux/huge_mm.h (1 line)

Commandline:
spatch --sp-file mmap_sem.cocci --dir . --macro-file coccinelle.h

If I explicitly pass the two headerfiles as parameters, they get handled
as well.  Maybe a bug in the recursive directory walk?

arch/mips/mm/fault.c seems to have a problem with one of the
conditionals.  I could reduce things to the following testcase:

void foo(unsigned long address)
{
        down_read(&mm->mmap_sem);
        vma = find_vma(current->mm, address);
        if (!vma)
                goto bad_area;
        if (vma->vm_start <= address)
                goto good_area;
        if (!(vma->vm_flags & VM_GROWSDOWN))
                goto bad_area;
        if (expand_stack(vma, address))
                goto bad_area;
bad_area:
        up_read(&mm->mmap_sem);
}

Remove the "!(vma->vm_flags & VM_GROWSDOWN)" condition and it works.
Leave the condition and spatch fails silently.

> > I run into problems whenever handling system calls.  Looks like the
> > SYSCALL_DEFINE5() macro and friends are causing problems.  Here is my
> > attempt to handle things.
> > 
> > #define SYSCALL_DEFINE1(func, t1, a1) \
> >     asmlinkage unsigned long func(t1 a1)
> > #define SYSCALL_DEFINE2(func, t1, a1, t2, a2) \
> >     asmlinkage unsigned long func(t1 a1, t2 a2)
> > #define SYSCALL_DEFINE3(func, t1, a1, t2, a2, t3, a3) \
> >     asmlinkage unsigned long func(t1 a1, t2 a2, t3 a3)
> > #define SYSCALL_DEFINE4(func, t1, a1, t2, a2, t3, a3, t4, a4) \
> >     asmlinkage unsigned long func(t1 a1, t2 a2, t3 a3, t4 a4)
> > #define SYSCALL_DEFINE5(func, t1, a1, t2, a2, t3, a3, t4, a4, t5, a5) \
> >     asmlinkage unsigned long func(t1 a1, t2 a2, t3 a3, t4 a4, t5 a5)
> > #define SYSCALL_DEFINE6(func, t1, a1, t2, a2, t3, a3, t4, a4, t5, a5, t6, 
> > a6) \
> >     asmlinkage unsigned long func(t1 a1, t2 a2, t3 a3, t4 a4, t5 a5, t6 a6)
> > 
> > Should that be included into standard.h?
> 
> If it works, I could certainly add it.

It works.

And tracking down failures in coccinelle is definitely more fun than
making 581 manual changes without messing things up.  Thank you!

Jörn

--
Homo Sapiens is a goal, not a description.
-- unknown
_______________________________________________
Cocci mailing list
[email protected]
https://systeme.lip6.fr/mailman/listinfo/cocci

Reply via email to