Skip to contentSkip to conversation

Replacing combination of kcalloc() and copy_from_user() by memdup_user() with SmPL?

Cocci SF Markus Elfring 2 participants 4 messages Feb 10, 2017 – Feb 15, 2017

lore
Share & tools
lore
#0 lore raw

Hello,

I came along another function implementation where I got the impression that the function ?memdup_user? could also be used there in a way which is similar to an existing source code transformation approach by a small script for the semantic patch language.

SmPL script:
https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/tree/scripts/coccinelle/api/memdup_user.cocci?id=632571b1bee00494aef749512d9f3290dfba0ead#n1

@refactoring@
expression code, count, from, to, size;
identifier label, log;
@@
to =
- kcalloc(count, size, GFP_KERNEL)
+ memdup_user(from, size * count)
;
if (
- !to
+ IS_ERR(to)

                || ...) {
-   code = -ENOMEM;
-   goto label;
-}
-code = copy_from_user(to, from, (size) * (count));
-if (code) {
    ...
    log(...,
-       code
+       PTR_ERR(to)
        , ...);
    code = -ENOMEM;
    goto label;

}

Source file:
https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git/tree/drivers/infiniband/hw/hfi1/user_sdma.c?id=632571b1bee00494aef749512d9f3290dfba0ead#n718

int hfi1_user_sdma_process_request(struct file *fp, struct iovec *iovec,
unsigned long dim, unsigned long *count)

{
	/*
	 * ...
	 */
	if (req_opcode(req->info.ctrl) == EXPECTED) {
		u16 ntids = iovec[idx].iov_len / sizeof(*req->tids);

		if (!ntids || ntids > MAX_TID_PAIR_ENTRIES) {
			ret = -EINVAL;
			goto free_req;
		}
		req->tids = kcalloc(ntids, sizeof(*req->tids), GFP_KERNEL);
		if (!req->tids) {
			ret = -ENOMEM;
			goto free_req;
		}
		/*
		 * ...
		 */
		ret = copy_from_user(req->tids, iovec[idx].iov_base,
				     ntids * sizeof(*req->tids));
		if (ret) {
			SDMA_DBG(req, "Failed to copy %d TIDs (%d)",
				 ntids, ret);
			ret = -EFAULT;
			goto free_req;
		}
		req->n_tids = ntids;
		idx++;
	}
	/*
	 * ...
	 */
	return 0;
free_req:
	user_sdma_free_request(req, true);
	if (req_queued)
		pq_update(pq);
	set_comp_state(pq, cq, info.comp_idx, ERROR, req->status);
	return ret;
}

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt replace_kcalloc_by_memdup_user-draft1.cocci ../Probe/user_sdma-excerpt1.c ?
refactoring: position variables or mixed modifs interfere with comm_assoc iso
bool (bool !to

Quoted 1 line
Quoted 1 line
Quoted 1 line

IS_ERR(to)

|| ...)
?

The software combination ?Coccinelle 1.0.6-00093-ge1776d7b (OCaml 4.03.0)? does not like this change attempt so far.
Which details should I reconsider next?

Regards,
Markus

#1 lore raw
Quoted 7 lines

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt replace_kcalloc_by_memdup_user-draft1.cocci ../Probe/user_sdma-excerpt1.c ?
refactoring: position variables or mixed modifs interfere with comm_assoc iso
bool (bool !to

Quoted 1 line
Quoted 1 line
Quoted 1 line

IS_ERR(to)

|| ...)
?

Would you like to explain this message a bit more while the following SmPL approach seems to work as expected?

@refactoring@
expression code, count, from, to, size;
identifier label, log;
@@
to =
- kcalloc(count, size, GFP_KERNEL)
+ memdup_user(from, size * count)
;
if (
- !to
+ IS_ERR(to)

    ) {

- code = -ENOMEM;
- goto label;
-}
-code = copy_from_user(to, from, (size) * (count));

-if (code) {
    ...
+   code = PTR_ERR(to);
    log(..., code, ...);
-   code = -EFAULT;
    goto label;

}

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt replace_kcalloc_by_memdup_user-draft2.cocci ../Probe/user_sdma-excerpt1.c
init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
HANDLING: ../Probe/user_sdma-excerpt1.c

tmp/cocci-output-8749-f5da06-user_sdma-excerpt1.c renamed from Probe/user_sdma-excerpt1.c +4 -12
diff = 
@@ -11,20 +11,12 @@ int hfi1_user_sdma_process_request(struc
 			ret = -EINVAL;
 			goto free_req;
 		}
-		req->tids = kcalloc(ntids, sizeof(*req->tids), GFP_KERNEL);
-		if (!req->tids) {
-			ret = -ENOMEM;
-			goto free_req;
-		}
-		/*
-		 * ...
-		 */
-		ret = copy_from_user(req->tids, iovec[idx].iov_base,
-				     ntids * sizeof(*req->tids));
-		if (ret) {
+		req->tids = memdup_user(iovec[idx].iov_base,
+					sizeof(*req->tids) * ntids);
+		if (IS_ERR(req->tids)) {
+			ret = PTR_ERR(req->tids);
 			SDMA_DBG(req, "Failed to copy %d TIDs (%d)",
 				 ntids, ret);
-			ret = -EFAULT;
 			goto free_req;
 		}
 		req->n_tids = ntids;

Regards,
Markus

#2 lore raw
Quoted 10 lines · SF Markus Elfring
Quoted 7 lines

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt replace_kcalloc_by_memdup_user-draft1.cocci ../Probe/user_sdma-excerpt1.c ?
refactoring: position variables or mixed modifs interfere with comm_assoc iso
bool (bool !to

Quoted 1 line
Quoted 1 line
Quoted 1 line

IS_ERR(to)

|| ...)
?

Would you like to explain this message a bit more while the following SmPL approach seems to work as expected?

The comm_assoc isomorphism allows A || ... to match any disjunction where A appears at the top level, not just in the far left position. It also matches the case where A is not present at all.

I don't know what the original semantic patch looks like, so I don't know exactly what is preventing this isomorphism from applying.

julia

Quoted 66 lines

@refactoring@
expression code, count, from, to, size;
identifier label, log;
@@
to =

(quoted diff) +3 -8
-     kcalloc(count, size, GFP_KERNEL)
+     memdup_user(from, size * count)
 ;
 if (
-    !to
+    IS_ERR(to)
    ) {
-   code = -ENOMEM;
-   goto label;
-}
-code = copy_from_user(to, from, (size) * (count));
-if (code) {
...
+   code = PTR_ERR(to);
    log(..., code, ...);
-   code = -EFAULT;
    goto label;
 }

elfring at Sonne:~/Projekte/Coccinelle/janitor> spatch.opt replace_kcalloc_by_memdup_user-draft2.cocci ../Probe/user_sdma-excerpt1.c
init_defs_builtins: /usr/local/lib64/coccinelle/standard.h
HANDLING: ../Probe/user_sdma-excerpt1.c
diff =

tmp/cocci-output-8749-f5da06-user_sdma-excerpt1.c renamed from Probe/user_sdma-excerpt1.c +4 -12
@@ -11,20 +11,12 @@ int hfi1_user_sdma_process_request(struc
 			ret = -EINVAL;
 			goto free_req;
 		}
-		req->tids = kcalloc(ntids, sizeof(*req->tids), GFP_KERNEL);
-		if (!req->tids) {
-			ret = -ENOMEM;
-			goto free_req;
-		}
-		/*
-		 * ...
-		 */
-		ret = copy_from_user(req->tids, iovec[idx].iov_base,
-				     ntids * sizeof(*req->tids));
-		if (ret) {
+		req->tids = memdup_user(iovec[idx].iov_base,
+					sizeof(*req->tids) * ntids);
+		if (IS_ERR(req->tids)) {
+			ret = PTR_ERR(req->tids);
 			SDMA_DBG(req, "Failed to copy %d TIDs (%d)",
 				 ntids, ret);
-			ret = -EFAULT;
 			goto free_req;
 		}
 		req->n_tids = ntids;

Regards,
Markus
_______________________________________________
Cocci mailing list
Cocci at systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci

#3 lore raw
Quoted 2 lines

The comm_assoc isomorphism allows A || ... to match any disjunction where A appears at the top level, not just in the far left position.

Thanks for your explanation.

Are there further software development challenges to consider around the handling of items like position variables and ?mixed modifs??

Quoted 1 line

It also matches the case where A is not present at all.

Will the bug report ?Safer source code analysis by "memdup_user.cocci"? get another look?
https://github.com/coccinelle/coccinelle/issues/78

Regards,
Markus