[v2] tty: rfcomm: prefer struct_size over open coded arithmetic

Commit Message

Erick Archer May 12, 2024, 11:17 a.m. UTC

This is an effort to get rid of all multiplications from allocation
functions in order to prevent integer overflows [1][2].

As the "dl" variable is a pointer to "struct rfcomm_dev_list_req" and
this structure ends in a flexible array:

struct rfcomm_dev_list_req {
	[...]
	struct   rfcomm_dev_info dev_info[];
};

the preferred way in the kernel is to use the struct_size() helper to
do the arithmetic instead of the calculation "size + count * size" in
the kzalloc() and copy_to_user() functions.

At the same time, prepare for the coming implementation by GCC and Clang
of the __counted_by attribute. Flexible array members annotated with
__counted_by can have their accesses bounds-checked at run-time via
CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE (for
strcpy/memcpy-family functions).

In this case, it is important to note that the logic needs a little
refactoring to ensure that the "dev_num" member is initialized before
the first access to the flex array. Specifically, add the assignment
before the list_for_each_entry() loop.

Also remove the "size" variable as it is no longer needed and refactor
the list_for_each_entry() loop to use di[n] instead of (di + n).

This way, the code is more readable, idiomatic and safer.

This code was detected with the help of Coccinelle, and audited and
modified manually.

Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#open-coded-arithmetic-in-allocator-arguments [1]
Link: https://github.com/KSPP/linux/issues/160 [2]
Signed-off-by: Erick Archer <[e-mail]>
---
Changes in v2:
- Add the __counted_by() attribute (Kees Cook).
- Refactor the list_for_each_entry() loop to use di[n] instead of
  (di + n) (Kees Cook).

Previous versions:
v1 -> https://lore.kernel.org/linux-hardening/[e-mail]/
---
 include/net/bluetooth/rfcomm.h |  2 +-
 net/bluetooth/rfcomm/tty.c     | 23 ++++++++++-------------
 2 files changed, 11 insertions(+), 14 deletions(-)

Comments

[e-mail] May 12, 2024, 11:55 a.m. UTC | #1

This is automated email and please do not reply to this email!

Dear submitter,

Thank you for submitting the patches to the linux bluetooth mailing list.
This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/project/bluetooth/list/?series=852638

---Test result---

Test Summary:
CheckPatch                    PASS      0.83 seconds
GitLint                       FAIL      0.47 seconds
SubjectPrefix                 FAIL      0.31 seconds
BuildKernel                   PASS      29.53 seconds
CheckAllWarning               PASS      32.27 seconds
CheckSparse                   PASS      38.11 seconds
CheckSmatch                   FAIL      35.80 seconds
BuildKernel32                 PASS      28.62 seconds
TestRunnerSetup               PASS      517.15 seconds
TestRunner_l2cap-tester       PASS      18.47 seconds
TestRunner_iso-tester         FAIL      31.31 seconds
TestRunner_bnep-tester        PASS      4.69 seconds
TestRunner_mgmt-tester        PASS      111.49 seconds
TestRunner_rfcomm-tester      PASS      7.21 seconds
TestRunner_sco-tester         PASS      15.40 seconds
TestRunner_ioctl-tester       PASS      7.58 seconds
TestRunner_mesh-tester        PASS      5.75 seconds
TestRunner_smp-tester         PASS      6.70 seconds
TestRunner_userchan-tester    PASS      4.87 seconds
IncrementalBuild              PASS      27.48 seconds

Details
##############################
Test: GitLint - FAIL
Desc: Run gitlint
Output:
[v2] tty: rfcomm: prefer struct_size over open coded arithmetic

WARNING: I3 - ignore-body-lines: gitlint will be switching from using Python regex 'match' (match beginning) to 'search' (match anywhere) semantics. Please review your ignore-body-lines.regex option accordingly. To remove this warning, set general.regex-style-search=True. More details: https://jorisroovers.github.io/gitlint/configuration/#regex-style-search
10: B3 Line contains hard tab characters (\t): "	[...]"
11: B3 Line contains hard tab characters (\t): "	struct   rfcomm_dev_info dev_info[];"
37: B1 Line exceeds max length (117>80): "Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#open-coded-arithmetic-in-allocator-arguments [1]"
47: B1 Line exceeds max length (125>80): "v1 -> https://lore.kernel.org/linux-hardening/[e-mail]/"
##############################
Test: SubjectPrefix - FAIL
Desc: Check subject contains "Bluetooth" prefix
Output:
"Bluetooth: " prefix is not specified in the subject
##############################
Test: CheckSmatch - FAIL
Desc: Run smatch tool with source
Output:

Segmentation fault (core dumped)
make[4]: *** [scripts/Makefile.build:244: net/bluetooth/hci_core.o] Error 139
make[4]: *** Deleting file 'net/bluetooth/hci_core.o'
make[3]: *** [scripts/Makefile.build:485: net/bluetooth] Error 2
make[2]: *** [scripts/Makefile.build:485: net] Error 2
make[2]: *** Waiting for unfinished jobs....
Segmentation fault (core dumped)
make[4]: *** [scripts/Makefile.build:244: drivers/bluetooth/bcm203x.o] Error 139
make[4]: *** Deleting file 'drivers/bluetooth/bcm203x.o'
make[4]: *** Waiting for unfinished jobs....
Segmentation fault (core dumped)
make[4]: *** [scripts/Makefile.build:244: drivers/bluetooth/bpa10x.o] Error 139
make[4]: *** Deleting file 'drivers/bluetooth/bpa10x.o'
make[3]: *** [scripts/Makefile.build:485: drivers/bluetooth] Error 2
make[2]: *** [scripts/Makefile.build:485: drivers] Error 2
make[1]: *** [/github/workspace/src/src/Makefile:1919: .] Error 2
make: *** [Makefile:240: __sub-make] Error 2
##############################
Test: TestRunner_iso-tester - FAIL
Desc: Run iso-tester with test-runner
Output:
Total: 122, Passed: 121 (99.2%), Failed: 1, Not Run: 0

Failed Test Cases
ISO Connect Suspend - Success                        Failed       4.172 seconds


---
Regards,
Linux Bluetooth

Kees Cook May 13, 2024, 3:03 a.m. UTC | #2

On Sun, May 12, 2024 at 01:17:24PM +0200, Erick Archer wrote:
> This is an effort to get rid of all multiplications from allocation
> functions in order to prevent integer overflows [1][2].
> 
> As the "dl" variable is a pointer to "struct rfcomm_dev_list_req" and
> this structure ends in a flexible array:
> 
> struct rfcomm_dev_list_req {
> 	[...]
> 	struct   rfcomm_dev_info dev_info[];
> };
> 
> the preferred way in the kernel is to use the struct_size() helper to
> do the arithmetic instead of the calculation "size + count * size" in
> the kzalloc() and copy_to_user() functions.
> 
> At the same time, prepare for the coming implementation by GCC and Clang
> of the __counted_by attribute. Flexible array members annotated with
> __counted_by can have their accesses bounds-checked at run-time via
> CONFIG_UBSAN_BOUNDS (for array indexing) and CONFIG_FORTIFY_SOURCE (for
> strcpy/memcpy-family functions).
> 
> In this case, it is important to note that the logic needs a little
> refactoring to ensure that the "dev_num" member is initialized before
> the first access to the flex array. Specifically, add the assignment
> before the list_for_each_entry() loop.
> 
> Also remove the "size" variable as it is no longer needed and refactor
> the list_for_each_entry() loop to use di[n] instead of (di + n).
> 
> This way, the code is more readable, idiomatic and safer.
> 
> This code was detected with the help of Coccinelle, and audited and
> modified manually.
> 
> Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#open-coded-arithmetic-in-allocator-arguments [1]
> Link: https://github.com/KSPP/linux/issues/160 [2]
> Signed-off-by: Erick Archer <[e-mail]>

Looks good!

Reviewed-by: Kees Cook <[e-mail]>

> [...]
> -		bacpy(&(di + n)->src, &dev->src);
> -		bacpy(&(di + n)->dst, &dev->dst);
> +		bacpy(&di[n].src, &dev->src);
> +		bacpy(&di[n].dst, &dev->dst);

Not an issue with your patch, but this helper is really pointless in the
Bluetooth tree:

static inline void bacpy(bdaddr_t *dst, const bdaddr_t *src)
{
        memcpy(dst, src, sizeof(bdaddr_t));
}

So the above could just be:

		di[n].src = dev->src;
		di[n].dst = dev->dst;

:P

-Kees

Jiri Slaby May 13, 2024, 5:07 a.m. UTC | #3

On 12. 05. 24, 13:17, Erick Archer wrote:
> This is an effort to get rid of all multiplications from allocation
> functions in order to prevent integer overflows [1][2].
> 
> As the "dl" variable is a pointer to "struct rfcomm_dev_list_req" and
> this structure ends in a flexible array:
...
> --- a/include/net/bluetooth/rfcomm.h
> +++ b/include/net/bluetooth/rfcomm.h
...
> @@ -528,12 +527,12 @@ static int rfcomm_get_dev_list(void __user *arg)
>   	list_for_each_entry(dev, &rfcomm_dev_list, list) {
>   		if (!tty_port_get(&dev->port))
>   			continue;
> -		(di + n)->id      = dev->id;
> -		(di + n)->flags   = dev->flags;
> -		(di + n)->state   = dev->dlc->state;
> -		(di + n)->channel = dev->channel;
> -		bacpy(&(di + n)->src, &dev->src);
> -		bacpy(&(di + n)->dst, &dev->dst);
> +		di[n].id      = dev->id;
> +		di[n].flags   = dev->flags;
> +		di[n].state   = dev->dlc->state;
> +		di[n].channel = dev->channel;
> +		bacpy(&di[n].src, &dev->src);
> +		bacpy(&di[n].dst, &dev->dst);

This does not relate much to "prefer struct_size over open coded 
arithmetic". It should have been in a separate patch.

Other than that, LGTM.

thanks,

Luiz Augusto von Dentz May 13, 2024, 4:29 p.m. UTC | #4

Hi Jiri, Eric,

On Mon, May 13, 2024 at 1:07 AM Jiri Slaby <[e-mail]> wrote:
>
> On 12. 05. 24, 13:17, Erick Archer wrote:
> > This is an effort to get rid of all multiplications from allocation
> > functions in order to prevent integer overflows [1][2].
> >
> > As the "dl" variable is a pointer to "struct rfcomm_dev_list_req" and
> > this structure ends in a flexible array:
> ...
> > --- a/include/net/bluetooth/rfcomm.h
> > +++ b/include/net/bluetooth/rfcomm.h
> ...
> > @@ -528,12 +527,12 @@ static int rfcomm_get_dev_list(void __user *arg)
> >       list_for_each_entry(dev, &rfcomm_dev_list, list) {
> >               if (!tty_port_get(&dev->port))
> >                       continue;
> > -             (di + n)->id      = dev->id;
> > -             (di + n)->flags   = dev->flags;
> > -             (di + n)->state   = dev->dlc->state;
> > -             (di + n)->channel = dev->channel;
> > -             bacpy(&(di + n)->src, &dev->src);
> > -             bacpy(&(di + n)->dst, &dev->dst);
> > +             di[n].id      = dev->id;
> > +             di[n].flags   = dev->flags;
> > +             di[n].state   = dev->dlc->state;
> > +             di[n].channel = dev->channel;
> > +             bacpy(&di[n].src, &dev->src);
> > +             bacpy(&di[n].dst, &dev->dst);
>
> This does not relate much to "prefer struct_size over open coded
> arithmetic". It should have been in a separate patch.

+1, please split these changes into its own patch so we can apply it separately.

> Other than that, LGTM.
>
> thanks,
> --
> js
> suse labs
>

Erick Archer May 13, 2024, 5:12 p.m. UTC | #5

Hi Kees, Jiri and Luiz,
First of all, thanks for the reviews.

On Mon, May 13, 2024 at 12:29:04PM -0400, Luiz Augusto von Dentz wrote:
> Hi Jiri, Eric,
> 
> On Mon, May 13, 2024 at 1:07 AM Jiri Slaby <[e-mail]> wrote:
> >
> > On 12. 05. 24, 13:17, Erick Archer wrote:
> > > This is an effort to get rid of all multiplications from allocation
> > > functions in order to prevent integer overflows [1][2].
> > >
> > > As the "dl" variable is a pointer to "struct rfcomm_dev_list_req" and
> > > this structure ends in a flexible array:
> > ...
> > > --- a/include/net/bluetooth/rfcomm.h
> > > +++ b/include/net/bluetooth/rfcomm.h
> > ...
> > > @@ -528,12 +527,12 @@ static int rfcomm_get_dev_list(void __user *arg)
> > >       list_for_each_entry(dev, &rfcomm_dev_list, list) {
> > >               if (!tty_port_get(&dev->port))
> > >                       continue;
> > > -             (di + n)->id      = dev->id;
> > > -             (di + n)->flags   = dev->flags;
> > > -             (di + n)->state   = dev->dlc->state;
> > > -             (di + n)->channel = dev->channel;
> > > -             bacpy(&(di + n)->src, &dev->src);
> > > -             bacpy(&(di + n)->dst, &dev->dst);
> > > +             di[n].id      = dev->id;
> > > +             di[n].flags   = dev->flags;
> > > +             di[n].state   = dev->dlc->state;
> > > +             di[n].channel = dev->channel;
> > > +             bacpy(&di[n].src, &dev->src);
> > > +             bacpy(&di[n].dst, &dev->dst);
> >
> > This does not relate much to "prefer struct_size over open coded
> > arithmetic". It should have been in a separate patch.
> 
> +1, please split these changes into its own patch so we can apply it separately.

Ok, no problem. Also, I will simplify the "bacpy" lines with direct
assignments as Kees suggested:

   di[n].src = dev->src;
   di[n].dst = dev->dst;

instead of:

   bacpy(&di[n].src, &dev->src);
   bacpy(&di[n].dst, &dev->dst);

Regards,
Erick

> > Other than that, LGTM.
> >
> > thanks,
> > --
> > js
> > suse labs
> >
> 
> 
> -- 
> Luiz Augusto von Dentz

Kees Cook May 13, 2024, 6:51 p.m. UTC | #6

On Mon, May 13, 2024 at 07:12:57PM +0200, Erick Archer wrote:
> Hi Kees, Jiri and Luiz,
> First of all, thanks for the reviews.
> 
> On Mon, May 13, 2024 at 12:29:04PM -0400, Luiz Augusto von Dentz wrote:
> > Hi Jiri, Eric,
> > 
> > On Mon, May 13, 2024 at 1:07 AM Jiri Slaby <[e-mail]> wrote:
> > >
> > > On 12. 05. 24, 13:17, Erick Archer wrote:
> > > > This is an effort to get rid of all multiplications from allocation
> > > > functions in order to prevent integer overflows [1][2].
> > > >
> > > > As the "dl" variable is a pointer to "struct rfcomm_dev_list_req" and
> > > > this structure ends in a flexible array:
> > > ...
> > > > --- a/include/net/bluetooth/rfcomm.h
> > > > +++ b/include/net/bluetooth/rfcomm.h
> > > ...
> > > > @@ -528,12 +527,12 @@ static int rfcomm_get_dev_list(void __user *arg)
> > > >       list_for_each_entry(dev, &rfcomm_dev_list, list) {
> > > >               if (!tty_port_get(&dev->port))
> > > >                       continue;
> > > > -             (di + n)->id      = dev->id;
> > > > -             (di + n)->flags   = dev->flags;
> > > > -             (di + n)->state   = dev->dlc->state;
> > > > -             (di + n)->channel = dev->channel;
> > > > -             bacpy(&(di + n)->src, &dev->src);
> > > > -             bacpy(&(di + n)->dst, &dev->dst);
> > > > +             di[n].id      = dev->id;
> > > > +             di[n].flags   = dev->flags;
> > > > +             di[n].state   = dev->dlc->state;
> > > > +             di[n].channel = dev->channel;
> > > > +             bacpy(&di[n].src, &dev->src);
> > > > +             bacpy(&di[n].dst, &dev->dst);
> > >
> > > This does not relate much to "prefer struct_size over open coded
> > > arithmetic". It should have been in a separate patch.
> > 
> > +1, please split these changes into its own patch so we can apply it separately.
> 
> Ok, no problem. Also, I will simplify the "bacpy" lines with direct
> assignments as Kees suggested:
> 
>    di[n].src = dev->src;
>    di[n].dst = dev->dst;
> 
> instead of:
> 
>    bacpy(&di[n].src, &dev->src);
>    bacpy(&di[n].dst, &dev->dst);

I think that's a separate thing and you can leave bacpy() as-is for now.

Patch

diff --git a/include/net/bluetooth/rfcomm.h b/include/net/bluetooth/rfcomm.h
index 99d26879b02a..c05882476900 100644
--- a/include/net/bluetooth/rfcomm.h
+++ b/include/net/bluetooth/rfcomm.h
@@ -355,7 +355,7 @@  struct rfcomm_dev_info {
 
 struct rfcomm_dev_list_req {
 	u16      dev_num;
-	struct   rfcomm_dev_info dev_info[];
+	struct   rfcomm_dev_info dev_info[] __counted_by(dev_num);
 };
 
 int  rfcomm_dev_ioctl(struct sock *sk, unsigned int cmd, void __user *arg);
diff --git a/net/bluetooth/rfcomm/tty.c b/net/bluetooth/rfcomm/tty.c
index 69c75c041fe1..af80d599c337 100644
--- a/net/bluetooth/rfcomm/tty.c
+++ b/net/bluetooth/rfcomm/tty.c
@@ -504,7 +504,7 @@  static int rfcomm_get_dev_list(void __user *arg)
 	struct rfcomm_dev *dev;
 	struct rfcomm_dev_list_req *dl;
 	struct rfcomm_dev_info *di;
-	int n = 0, size, err;
+	int n = 0, err;
 	u16 dev_num;
 
 	BT_DBG("");
@@ -515,12 +515,11 @@  static int rfcomm_get_dev_list(void __user *arg)
 	if (!dev_num || dev_num > (PAGE_SIZE * 4) / sizeof(*di))
 		return -EINVAL;
 
-	size = sizeof(*dl) + dev_num * sizeof(*di);
-
-	dl = kzalloc(size, GFP_KERNEL);
+	dl = kzalloc(struct_size(dl, dev_info, dev_num), GFP_KERNEL);
 	if (!dl)
 		return -ENOMEM;
 
+	dl->dev_num = dev_num;
 	di = dl->dev_info;
 
 	mutex_lock(&rfcomm_dev_lock);
@@ -528,12 +527,12 @@  static int rfcomm_get_dev_list(void __user *arg)
 	list_for_each_entry(dev, &rfcomm_dev_list, list) {
 		if (!tty_port_get(&dev->port))
 			continue;
-		(di + n)->id      = dev->id;
-		(di + n)->flags   = dev->flags;
-		(di + n)->state   = dev->dlc->state;
-		(di + n)->channel = dev->channel;
-		bacpy(&(di + n)->src, &dev->src);
-		bacpy(&(di + n)->dst, &dev->dst);
+		di[n].id      = dev->id;
+		di[n].flags   = dev->flags;
+		di[n].state   = dev->dlc->state;
+		di[n].channel = dev->channel;
+		bacpy(&di[n].src, &dev->src);
+		bacpy(&di[n].dst, &dev->dst);
 		tty_port_put(&dev->port);
 		if (++n >= dev_num)
 			break;
@@ -542,9 +541,7 @@  static int rfcomm_get_dev_list(void __user *arg)
 	mutex_unlock(&rfcomm_dev_lock);
 
 	dl->dev_num = n;
-	size = sizeof(*dl) + n * sizeof(*di);
-
-	err = copy_to_user(arg, dl, size);
+	err = copy_to_user(arg, dl, struct_size(dl, dev_info, n));
 	kfree(dl);
 
 	return err ? -EFAULT : 0;