Skip to contentSkip to conversation

Coccinelle: pci_free_consistent: Extend whenconstraints for two SmPL ellipses

PATCH Julia Lawall 2 participants 9 messages May 15, 2019 2 messages not in the archive

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw
Quoted 13 lines · On Wed, 15 May 2019, [email protected] wrote:

Hello,

Quoted 12 lines
Quoted 11 lines

A SmPL ellipsis was specified for a search approach so that additional source code would be tolerated between an assignment to a local variable and the corresponding null pointer check.

But such code should be restricted.
* The local variable must not be reassigned there.
* It must also not be forwarded to an other assignment target.

Take additional casts for these code exclusion specifications into account together with optional parentheses.

Quoted 6 lines
Quoted 3 lines

I leave this up to the ZTE people.
julia

Thanks.

1, "id = (T2)(e)" is rare.

Thanks for checking. I don't really care if it is rare. There should not be much cost to this. On the other hand, I do care about causing false negatives. I don't know any more what is the type of id. Making it identifier would lead to false negatives as noted below. It would be better as something like e2->fld.

julia

Quoted 66 lines

It may be a minor detail that will have no impact in practice.
We've tested it, and this SmPL may only need to fix the following two false positives:
./drivers/net/ethernet/cisco/enic/vnic_dev.c:861:1-7: ERROR: missing pci_free_consistent; pci_alloc_consistent on line 855 and return without freeing on line 861 ./drivers/message/fusion/mptctl.c:2643:2-8: ERROR: missing pci_free_consistent; pci_alloc_consistent on line 2511 and return without freeing on line 2643

2, If you really plan to add the two restrictions above,
you may need to consider this further than simply adding a "when != id = (T2)(e)" statement. I constructed the flollowing code snippet as a test case:

int test()

{
	unsigned char *new_page = NULL;
	int ret;

	new_page = pci_alloc_consistent(dev, 4096, &temp_pci);
	if (new_page == NULL)
    return -1;
X;
	Y;
	new_page = Z;

	return ret;
}

Using the original SmPL, we can find a bug. But with your modified SmPL, we can't find the bug.

--
Regards,
Wen

Quoted 35 lines
Quoted 35 lines
Fixes f7b167113753e95ae61383e234f8d10142782ace ("scripts: Coccinelle script for pci_free_consistent()")
---
 scripts/coccinelle/free/pci_free_consistent.cocci | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
scripts/coccinelle/free/pci_free_consistent.cocci +4 -3
index 45bc14ece151..48a36adfa3ce 100644
@@ -13,13 +13,15 @@ virtual org
 local idexpression id;
 expression x,y,z,e;
 position p1,p2;
-type T;
+type T,T2,T3,T4;
 @@
 
 id = pci_alloc_consistent@p1(x,y,&z)
-... when != e = id
+ ... when != id = (T2)(e)
+     when != e = (T3)(id)
 if (id == NULL || ...) { ... return ...; }
 ... when != pci_free_consistent(x,y,id,z)
+    when != id = (T4)(e)
     when != if (id) { ... pci_free_consistent(x,y,id,z) ... }
     when != if (y) { ... pci_free_consistent(x,y,id,z) ... }
     when != e = (T)id
--

2.21.0

#2 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

Quoted 3 lines
Quoted 1 line

1, "id = (T2)(e)" is rare.

Thanks for checking. I don't really care if it is rare.

I got related source code analysis concerns.

Quoted 1 line

There should not be much cost to this.

Each additional filter will influence the software run times and possible results.

Quoted 1 line

On the other hand, I do care about causing false negatives.

Do you find the missing warning after the addition of such an exclusion specification interesting?

Regards,
Markus

A message that is not in the archive (sent off-list or not yet received): alpine.DEB.2.20.1905151119070.3231@hadrien

#4 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

On the other hand, I do care about causing false negatives.

Do you find the missing warning after the addition of such an exclusion specification interesting?

I already suggested how to improve the code.

I find that the idea “e2->fld” needs further clarification. Such a SmPL specification will be resolved also to an expression, won't it?

I guess that it can fit better to other analysis approaches. Will any script variants become more interesting?

Is it still unclear which detail (besides the added SmPL filter) will hinder the warning display for Wen's test case?

Regards,
Markus

#5 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

Quoted 10 lines · Markus Elfring
Quoted 6 lines
Quoted 4 lines
Quoted 1 line

On the other hand, I do care about causing false negatives.

Do you find the missing warning after the addition of such an exclusion specification interesting?

I already suggested how to improve the code.

I find that the idea “e2->fld” needs further clarification. Such a SmPL specification will be resolved also to an expression, won't it?

Saving in a local variable doesn't impact the need to free the object. A field is the most obvious case where the object may not need freeing. But there are many expressions that e2->fld will not match.

julia

#6 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

Am 15.05.19 um 12:19 schrieb Julia Lawall:

Quoted 9 lines · Markus Elfring
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Quoted 2 lines

On the other hand, I do care about causing false negatives.

Do you find the missing warning after the addition of such an exclusion specification interesting?

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

I already suggested how to improve the code.

I find that the idea “e2->fld” needs further clarification. Such a SmPL specification will be resolved also to an expression, won't it?

Saving in a local variable doesn't impact the need to free the object.

I suggest to reconsider this view.

Would we like to introduce additional case distinctions for the handling of reassignments to local variables (as shown in Wen's test case)?

Quoted 1 line

A field is the most obvious case where the object may not need freeing.

A corresponding resource release should probably be performed by an other function then.

Quoted 1 line

But there are many expressions that e2->fld will not match.

Data structure members can eventually belong also to a local variable. Would they become relevant for further SmPL exclusion specifications?

Regards,
Markus

#7 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

In reply to a message not in the archive
Quoted 1 line

1, "id = (T2)(e)" is rare.

I can follow this view for such a filter.

Quoted 1 line

It may be a minor detail that will have no impact in practice.

I suggest to reconsider this view once more.

Should such exclusion specifications take also “unexpected” source code into account so that analysis results will be presented with a low false positive rate?

Quoted 1 line

We've tested it, and this SmPL may only need to fix the following two false positives:

Thanks for your acknowledgement that my proposal can give us another useful effect.

Quoted 3 lines

2, If you really plan to add the two restrictions above,
you may need to consider this further than simply adding a "when != id = (T2)(e)" statement. I constructed the flollowing code snippet as a test case:

…

Quoted 1 line

Using the original SmPL, we can find a bug.

I observe on my system that I do not get a desired warning
by the software combination “Coccinelle 1.0.7-00186-g99e081e9 (OCaml 4.07.1)” (even from the unmodified SmPL script) for your test example.

Which version are you using for the spatch program?

Quoted 1 line

But with your modified SmPL, we can't find the bug.

I do not see a difference here. - I wonder also about this situation then.

But this gives us the opportunity to clarify the really desired software behaviour in more detail.
How many developers would like to help with additional insights?

Regards,
Markus

#8 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

Quoted 3 lines
Quoted 1 line

But with your modified SmPL, we can't find the bug.

I do not see a difference here.

I have accidentally overlooked to pass the parameter “-D report”
to the spatch program.
Now I notice also an unexpected and undesirable difference for your test example after the suggested addition of exclusion specifications.

Quoted 1 line

- I wonder also about this situation then.

How should the software behaviour be improved finally?

Will we need to work with any SmPL script versions in parallel for selected source code analysis configurations?

Regards,
Markus

#9 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

In reply to a message not in the archive
Quoted 1 line

1, "id = (T2)(e)" is rare.

I can follow this view for such a filter.

Quoted 1 line

It may be a minor detail that will have no impact in practice.

I suggest to reconsider this view once more.

Should such exclusion specifications take also “unexpected” source code into account so that analysis results will be presented with a low false positive rate?

Quoted 1 line

We've tested it, and this SmPL may only need to fix the following two false positives:

Thanks for your acknowledgement that my proposal can give us another useful effect.

Quoted 3 lines

2, If you really plan to add the two restrictions above,
you may need to consider this further than simply adding a "when != id = (T2)(e)" statement. I constructed the flollowing code snippet as a test case:

…

Quoted 1 line

Using the original SmPL, we can find a bug.

I observe on my system that I do not get a desired warning
by the software combination “Coccinelle 1.0.7-00186-g99e081e9 (OCaml 4.07.1)” (even from the unmodified SmPL script) for your test example.

Which version are you using for the spatch program?

Quoted 1 line

But with your modified SmPL, we can't find the bug.

I do not see a difference here. - I wonder also about this situation then.

But this gives us the opportunity to clarify the really desired software behaviour in more detail.
How many developers would like to help with additional insights?

Regards,
Markus

#10 lore raw

Re: [3/3] Coccinelle: pci_free_consistent: Extend when constraints for two SmPL ellipses

In reply to a message not in the archive
Quoted 3 lines

2, If you really plan to add the two restrictions above,
you may need to consider this further than simply adding a "when != id = (T2)(e)" statement. I constructed the flollowing code snippet as a test case:

…

Quoted 2 lines

Using the original SmPL, we can find a bug. But with your modified SmPL, we can't find the bug.

The difference is the possible exclusion of a reassignment to a local variable in which a pointer was stored for a resource allocation.

Would you like to adjust corresponding case distinctions any further?

How many script variants will become relevant here?

Regards,
Markus