* Re: [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
[not found] < <CAG48ez1DPWhT2dhd1iptFawWjteh_=pZ4M6Yq5KKCq2DTArnqw@mail.gmail.com>
@ 2025-05-02 12:41 ` David Laight
0 siblings, 0 replies; 7+ messages in thread
From: David Laight @ 2025-05-02 12:41 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 1112 bytes --]
On Fri, 2 May 2025 14:28:21 +0200
Jann Horn <jannh@google.com> wrote:
> On Thu, May 1, 2025 at 10:02 PM David Laight
> <david.laight.linux@gmail.com> wrote:
> >
> > On Mon, 28 Apr 2025 15:06:43 +0200
> > Jann Horn <jannh@google.com> wrote:
> >
> > > The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
> > > 32-bit kernels: Multiplying a user-provided 32-bit value with the
> > > size of a structure can wrap around on such platforms.
> > >
> > > Fix it by using saturating arithmetic for the size calculation.
> >
> > Why not just add a sanity check on 'num_params' after it is read.
> > Max is 31 (1024-32)/32), but any sane limit will do because of
> > the buf.buf_len test.
>
> That would work, too. I don't know which way looks nicer.
The saturating arithmetic functions are non-obvious and non-trivial.
I looked at the code to check where buf.buf_len came from,
without its sanity check the user could craft a request where it
matched the saturated size.
I think I'd sanity check the number of items and then check that that
buffer length is right for the number of items.
David
^ permalink raw reply [flat|nested] 7+ messages in thread
[parent not found: < <CAK8z29XWCujUdLXcHz075+xcix8HV2Mp3EtxxX9GB7vGjwi3HA@mail.gmail.com>]
* Re: [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
[not found] < <CAK8z29XWCujUdLXcHz075+xcix8HV2Mp3EtxxX9GB7vGjwi3HA@mail.gmail.com>
@ 2025-04-30 13:00 ` Jens Wiklander
0 siblings, 0 replies; 7+ messages in thread
From: Jens Wiklander @ 2025-04-30 13:00 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 4169 bytes --]
On Wed, Apr 30, 2025 at 2:12 PM Rouven Czerwinski
<rouven.czerwinski@linaro.org> wrote:
>
> Hi,
>
> On Wed, 30 Apr 2025 at 13:53, Jens Wiklander <jens.wiklander@linaro.org> wrote:
> >
> > On Mon, Apr 28, 2025 at 3:06 PM Jann Horn <jannh@google.com> wrote:
> > >
> > > The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
> > > 32-bit kernels: Multiplying a user-provided 32-bit value with the
> > > size of a structure can wrap around on such platforms.
> > >
> > > Fix it by using saturating arithmetic for the size calculation.
> > >
> > > This has no security consequences because, in all users of
> > > TEE_IOCTL_PARAM_SIZE(), the subsequent kcalloc() implicitly checks
> > > for wrapping.
> > >
> > > Signed-off-by: Jann Horn <jannh@google.com>
> > > ---
> > > Note that I don't have a test device with a TEE; I only compile-tested
> > > the change on an x86-64 build.
> > > ---
> > > drivers/tee/tee_core.c | 11 ++++++-----
> > > 1 file changed, 6 insertions(+), 5 deletions(-)
> >
> > Looks good, I'm picking up this.
> >
> > Thanks,
> > Jens
> >
> > >
> > > diff --git a/drivers/tee/tee_core.c b/drivers/tee/tee_core.c
> > > index d113679b1e2d..acc7998758ad 100644
> > > --- a/drivers/tee/tee_core.c
> > > +++ b/drivers/tee/tee_core.c
> > > @@ -10,6 +10,7 @@
> > > #include <linux/fs.h>
> > > #include <linux/idr.h>
> > > #include <linux/module.h>
> > > +#include <linux/overflow.h>
> > > #include <linux/slab.h>
> > > #include <linux/tee_core.h>
> > > #include <linux/uaccess.h>
> > > @@ -19,7 +20,7 @@
> > >
> > > #define TEE_NUM_DEVICES 32
> > >
> > > -#define TEE_IOCTL_PARAM_SIZE(x) (sizeof(struct tee_param) * (x))
> > > +#define TEE_IOCTL_PARAM_SIZE(x) (size_mul(sizeof(struct tee_param), (x)))
> > >
> > > #define TEE_UUID_NS_NAME_SIZE 128
> > >
> > > @@ -487,7 +488,7 @@ static int tee_ioctl_open_session(struct tee_context *ctx,
> > > if (copy_from_user(&arg, uarg, sizeof(arg)))
> > > return -EFAULT;
> > >
> > > - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> > > + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> > > return -EINVAL;
> > >
> > > if (arg.num_params) {
> > > @@ -565,7 +566,7 @@ static int tee_ioctl_invoke(struct tee_context *ctx,
> > > if (copy_from_user(&arg, uarg, sizeof(arg)))
> > > return -EFAULT;
> > >
> > > - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> > > + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> > > return -EINVAL;
> > >
> > > if (arg.num_params) {
> > > @@ -699,7 +700,7 @@ static int tee_ioctl_supp_recv(struct tee_context *ctx,
> > > if (get_user(num_params, &uarg->num_params))
> > > return -EFAULT;
> > >
> > > - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) != buf.buf_len)
> > > + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) != buf.buf_len)
> > > return -EINVAL;
> > >
> > > params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
> > > @@ -798,7 +799,7 @@ static int tee_ioctl_supp_send(struct tee_context *ctx,
> > > get_user(num_params, &uarg->num_params))
> > > return -EFAULT;
> > >
> > > - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) > buf.buf_len)
> > > + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) > buf.buf_len)
> > > return -EINVAL;
> > >
> > > params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
> > >
> > > ---
> > > base-commit: b4432656b36e5cc1d50a1f2dc15357543add530e
> > > change-id: 20250428-tee-sizecheck-299d5eff8fc7
> > >
> > > --
> > > Jann Horn <jannh@google.com>
> > >
>
> I ran this through the arm32 qemu virt machine to test my new development setup,
> so:
>
> Tested-by: Rouven Czerwinski <rouven.czerwinski@linaro.org>
Thanks for testing.
Cheers,
Jens
>
> Best regards,
> Rouven
^ permalink raw reply [flat|nested] 7+ messages in thread
[parent not found: < <CAHUa44E_JZdYnGrReP0zWCP1wdu2BdJ9DSZZ3a2OiobRj61ThQ@mail.gmail.com>]
* Re: [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
[not found] < <CAHUa44E_JZdYnGrReP0zWCP1wdu2BdJ9DSZZ3a2OiobRj61ThQ@mail.gmail.com>
@ 2025-04-30 12:12 ` Rouven Czerwinski
0 siblings, 0 replies; 7+ messages in thread
From: Rouven Czerwinski @ 2025-04-30 12:12 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 3830 bytes --]
Hi,
On Wed, 30 Apr 2025 at 13:53, Jens Wiklander <jens.wiklander@linaro.org> wrote:
>
> On Mon, Apr 28, 2025 at 3:06 PM Jann Horn <jannh@google.com> wrote:
> >
> > The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
> > 32-bit kernels: Multiplying a user-provided 32-bit value with the
> > size of a structure can wrap around on such platforms.
> >
> > Fix it by using saturating arithmetic for the size calculation.
> >
> > This has no security consequences because, in all users of
> > TEE_IOCTL_PARAM_SIZE(), the subsequent kcalloc() implicitly checks
> > for wrapping.
> >
> > Signed-off-by: Jann Horn <jannh@google.com>
> > ---
> > Note that I don't have a test device with a TEE; I only compile-tested
> > the change on an x86-64 build.
> > ---
> > drivers/tee/tee_core.c | 11 ++++++-----
> > 1 file changed, 6 insertions(+), 5 deletions(-)
>
> Looks good, I'm picking up this.
>
> Thanks,
> Jens
>
> >
> > diff --git a/drivers/tee/tee_core.c b/drivers/tee/tee_core.c
> > index d113679b1e2d..acc7998758ad 100644
> > --- a/drivers/tee/tee_core.c
> > +++ b/drivers/tee/tee_core.c
> > @@ -10,6 +10,7 @@
> > #include <linux/fs.h>
> > #include <linux/idr.h>
> > #include <linux/module.h>
> > +#include <linux/overflow.h>
> > #include <linux/slab.h>
> > #include <linux/tee_core.h>
> > #include <linux/uaccess.h>
> > @@ -19,7 +20,7 @@
> >
> > #define TEE_NUM_DEVICES 32
> >
> > -#define TEE_IOCTL_PARAM_SIZE(x) (sizeof(struct tee_param) * (x))
> > +#define TEE_IOCTL_PARAM_SIZE(x) (size_mul(sizeof(struct tee_param), (x)))
> >
> > #define TEE_UUID_NS_NAME_SIZE 128
> >
> > @@ -487,7 +488,7 @@ static int tee_ioctl_open_session(struct tee_context *ctx,
> > if (copy_from_user(&arg, uarg, sizeof(arg)))
> > return -EFAULT;
> >
> > - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> > + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> > return -EINVAL;
> >
> > if (arg.num_params) {
> > @@ -565,7 +566,7 @@ static int tee_ioctl_invoke(struct tee_context *ctx,
> > if (copy_from_user(&arg, uarg, sizeof(arg)))
> > return -EFAULT;
> >
> > - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> > + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> > return -EINVAL;
> >
> > if (arg.num_params) {
> > @@ -699,7 +700,7 @@ static int tee_ioctl_supp_recv(struct tee_context *ctx,
> > if (get_user(num_params, &uarg->num_params))
> > return -EFAULT;
> >
> > - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) != buf.buf_len)
> > + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) != buf.buf_len)
> > return -EINVAL;
> >
> > params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
> > @@ -798,7 +799,7 @@ static int tee_ioctl_supp_send(struct tee_context *ctx,
> > get_user(num_params, &uarg->num_params))
> > return -EFAULT;
> >
> > - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) > buf.buf_len)
> > + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) > buf.buf_len)
> > return -EINVAL;
> >
> > params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
> >
> > ---
> > base-commit: b4432656b36e5cc1d50a1f2dc15357543add530e
> > change-id: 20250428-tee-sizecheck-299d5eff8fc7
> >
> > --
> > Jann Horn <jannh@google.com>
> >
I ran this through the arm32 qemu virt machine to test my new development setup,
so:
Tested-by: Rouven Czerwinski <rouven.czerwinski@linaro.org>
Best regards,
Rouven
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
@ 2025-04-28 13:06 Jann Horn
2025-04-30 11:52 ` Jens Wiklander
2025-05-01 20:01 ` David Laight
0 siblings, 2 replies; 7+ messages in thread
From: Jann Horn @ 2025-04-28 13:06 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 2912 bytes --]
The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
32-bit kernels: Multiplying a user-provided 32-bit value with the
size of a structure can wrap around on such platforms.
Fix it by using saturating arithmetic for the size calculation.
This has no security consequences because, in all users of
TEE_IOCTL_PARAM_SIZE(), the subsequent kcalloc() implicitly checks
for wrapping.
Signed-off-by: Jann Horn <jannh@google.com>
---
Note that I don't have a test device with a TEE; I only compile-tested
the change on an x86-64 build.
---
drivers/tee/tee_core.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/drivers/tee/tee_core.c b/drivers/tee/tee_core.c
index d113679b1e2d..acc7998758ad 100644
--- a/drivers/tee/tee_core.c
+++ b/drivers/tee/tee_core.c
@@ -10,6 +10,7 @@
#include <linux/fs.h>
#include <linux/idr.h>
#include <linux/module.h>
+#include <linux/overflow.h>
#include <linux/slab.h>
#include <linux/tee_core.h>
#include <linux/uaccess.h>
@@ -19,7 +20,7 @@
#define TEE_NUM_DEVICES 32
-#define TEE_IOCTL_PARAM_SIZE(x) (sizeof(struct tee_param) * (x))
+#define TEE_IOCTL_PARAM_SIZE(x) (size_mul(sizeof(struct tee_param), (x)))
#define TEE_UUID_NS_NAME_SIZE 128
@@ -487,7 +488,7 @@ static int tee_ioctl_open_session(struct tee_context *ctx,
if (copy_from_user(&arg, uarg, sizeof(arg)))
return -EFAULT;
- if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
+ if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
return -EINVAL;
if (arg.num_params) {
@@ -565,7 +566,7 @@ static int tee_ioctl_invoke(struct tee_context *ctx,
if (copy_from_user(&arg, uarg, sizeof(arg)))
return -EFAULT;
- if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
+ if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
return -EINVAL;
if (arg.num_params) {
@@ -699,7 +700,7 @@ static int tee_ioctl_supp_recv(struct tee_context *ctx,
if (get_user(num_params, &uarg->num_params))
return -EFAULT;
- if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) != buf.buf_len)
+ if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) != buf.buf_len)
return -EINVAL;
params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
@@ -798,7 +799,7 @@ static int tee_ioctl_supp_send(struct tee_context *ctx,
get_user(num_params, &uarg->num_params))
return -EFAULT;
- if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) > buf.buf_len)
+ if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) > buf.buf_len)
return -EINVAL;
params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
---
base-commit: b4432656b36e5cc1d50a1f2dc15357543add530e
change-id: 20250428-tee-sizecheck-299d5eff8fc7
--
Jann Horn <jannh@google.com>
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
2025-04-28 13:06 Jann Horn
@ 2025-04-30 11:52 ` Jens Wiklander
2025-05-01 20:01 ` David Laight
1 sibling, 0 replies; 7+ messages in thread
From: Jens Wiklander @ 2025-04-30 11:52 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 3383 bytes --]
On Mon, Apr 28, 2025 at 3:06 PM Jann Horn <jannh@google.com> wrote:
>
> The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
> 32-bit kernels: Multiplying a user-provided 32-bit value with the
> size of a structure can wrap around on such platforms.
>
> Fix it by using saturating arithmetic for the size calculation.
>
> This has no security consequences because, in all users of
> TEE_IOCTL_PARAM_SIZE(), the subsequent kcalloc() implicitly checks
> for wrapping.
>
> Signed-off-by: Jann Horn <jannh@google.com>
> ---
> Note that I don't have a test device with a TEE; I only compile-tested
> the change on an x86-64 build.
> ---
> drivers/tee/tee_core.c | 11 ++++++-----
> 1 file changed, 6 insertions(+), 5 deletions(-)
Looks good, I'm picking up this.
Thanks,
Jens
>
> diff --git a/drivers/tee/tee_core.c b/drivers/tee/tee_core.c
> index d113679b1e2d..acc7998758ad 100644
> --- a/drivers/tee/tee_core.c
> +++ b/drivers/tee/tee_core.c
> @@ -10,6 +10,7 @@
> #include <linux/fs.h>
> #include <linux/idr.h>
> #include <linux/module.h>
> +#include <linux/overflow.h>
> #include <linux/slab.h>
> #include <linux/tee_core.h>
> #include <linux/uaccess.h>
> @@ -19,7 +20,7 @@
>
> #define TEE_NUM_DEVICES 32
>
> -#define TEE_IOCTL_PARAM_SIZE(x) (sizeof(struct tee_param) * (x))
> +#define TEE_IOCTL_PARAM_SIZE(x) (size_mul(sizeof(struct tee_param), (x)))
>
> #define TEE_UUID_NS_NAME_SIZE 128
>
> @@ -487,7 +488,7 @@ static int tee_ioctl_open_session(struct tee_context *ctx,
> if (copy_from_user(&arg, uarg, sizeof(arg)))
> return -EFAULT;
>
> - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> return -EINVAL;
>
> if (arg.num_params) {
> @@ -565,7 +566,7 @@ static int tee_ioctl_invoke(struct tee_context *ctx,
> if (copy_from_user(&arg, uarg, sizeof(arg)))
> return -EFAULT;
>
> - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> return -EINVAL;
>
> if (arg.num_params) {
> @@ -699,7 +700,7 @@ static int tee_ioctl_supp_recv(struct tee_context *ctx,
> if (get_user(num_params, &uarg->num_params))
> return -EFAULT;
>
> - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) != buf.buf_len)
> + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) != buf.buf_len)
> return -EINVAL;
>
> params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
> @@ -798,7 +799,7 @@ static int tee_ioctl_supp_send(struct tee_context *ctx,
> get_user(num_params, &uarg->num_params))
> return -EFAULT;
>
> - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) > buf.buf_len)
> + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) > buf.buf_len)
> return -EINVAL;
>
> params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
>
> ---
> base-commit: b4432656b36e5cc1d50a1f2dc15357543add530e
> change-id: 20250428-tee-sizecheck-299d5eff8fc7
>
> --
> Jann Horn <jannh@google.com>
>
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
2025-04-28 13:06 Jann Horn
2025-04-30 11:52 ` Jens Wiklander
@ 2025-05-01 20:01 ` David Laight
2025-05-02 12:28 ` Jann Horn
1 sibling, 1 reply; 7+ messages in thread
From: David Laight @ 2025-05-01 20:01 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 3280 bytes --]
On Mon, 28 Apr 2025 15:06:43 +0200
Jann Horn <jannh@google.com> wrote:
> The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
> 32-bit kernels: Multiplying a user-provided 32-bit value with the
> size of a structure can wrap around on such platforms.
>
> Fix it by using saturating arithmetic for the size calculation.
Why not just add a sanity check on 'num_params' after it is read.
Max is 31 (1024-32)/32), but any sane limit will do because of
the buf.buf_len test.
David
>
> This has no security consequences because, in all users of
> TEE_IOCTL_PARAM_SIZE(), the subsequent kcalloc() implicitly checks
> for wrapping.
>
> Signed-off-by: Jann Horn <jannh@google.com>
> ---
> Note that I don't have a test device with a TEE; I only compile-tested
> the change on an x86-64 build.
> ---
> drivers/tee/tee_core.c | 11 ++++++-----
> 1 file changed, 6 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/tee/tee_core.c b/drivers/tee/tee_core.c
> index d113679b1e2d..acc7998758ad 100644
> --- a/drivers/tee/tee_core.c
> +++ b/drivers/tee/tee_core.c
> @@ -10,6 +10,7 @@
> #include <linux/fs.h>
> #include <linux/idr.h>
> #include <linux/module.h>
> +#include <linux/overflow.h>
> #include <linux/slab.h>
> #include <linux/tee_core.h>
> #include <linux/uaccess.h>
> @@ -19,7 +20,7 @@
>
> #define TEE_NUM_DEVICES 32
>
> -#define TEE_IOCTL_PARAM_SIZE(x) (sizeof(struct tee_param) * (x))
> +#define TEE_IOCTL_PARAM_SIZE(x) (size_mul(sizeof(struct tee_param), (x)))
>
> #define TEE_UUID_NS_NAME_SIZE 128
>
> @@ -487,7 +488,7 @@ static int tee_ioctl_open_session(struct tee_context *ctx,
> if (copy_from_user(&arg, uarg, sizeof(arg)))
> return -EFAULT;
>
> - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> return -EINVAL;
>
> if (arg.num_params) {
> @@ -565,7 +566,7 @@ static int tee_ioctl_invoke(struct tee_context *ctx,
> if (copy_from_user(&arg, uarg, sizeof(arg)))
> return -EFAULT;
>
> - if (sizeof(arg) + TEE_IOCTL_PARAM_SIZE(arg.num_params) != buf.buf_len)
> + if (size_add(sizeof(arg), TEE_IOCTL_PARAM_SIZE(arg.num_params)) != buf.buf_len)
> return -EINVAL;
>
> if (arg.num_params) {
> @@ -699,7 +700,7 @@ static int tee_ioctl_supp_recv(struct tee_context *ctx,
> if (get_user(num_params, &uarg->num_params))
> return -EFAULT;
>
> - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) != buf.buf_len)
> + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) != buf.buf_len)
> return -EINVAL;
>
> params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
> @@ -798,7 +799,7 @@ static int tee_ioctl_supp_send(struct tee_context *ctx,
> get_user(num_params, &uarg->num_params))
> return -EFAULT;
>
> - if (sizeof(*uarg) + TEE_IOCTL_PARAM_SIZE(num_params) > buf.buf_len)
> + if (size_add(sizeof(*uarg), TEE_IOCTL_PARAM_SIZE(num_params)) > buf.buf_len)
> return -EINVAL;
>
> params = kcalloc(num_params, sizeof(struct tee_param), GFP_KERNEL);
>
> ---
> base-commit: b4432656b36e5cc1d50a1f2dc15357543add530e
> change-id: 20250428-tee-sizecheck-299d5eff8fc7
>
^ permalink raw reply [flat|nested] 7+ messages in thread* Re: [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels
2025-05-01 20:01 ` David Laight
@ 2025-05-02 12:28 ` Jann Horn
0 siblings, 0 replies; 7+ messages in thread
From: Jann Horn @ 2025-05-02 12:28 UTC (permalink / raw)
To: op-tee
[-- Attachment #1: Type: text/plain, Size: 652 bytes --]
On Thu, May 1, 2025 at 10:02 PM David Laight
<david.laight.linux@gmail.com> wrote:
>
> On Mon, 28 Apr 2025 15:06:43 +0200
> Jann Horn <jannh@google.com> wrote:
>
> > The current code around TEE_IOCTL_PARAM_SIZE() is a bit wrong on
> > 32-bit kernels: Multiplying a user-provided 32-bit value with the
> > size of a structure can wrap around on such platforms.
> >
> > Fix it by using saturating arithmetic for the size calculation.
>
> Why not just add a sanity check on 'num_params' after it is read.
> Max is 31 (1024-32)/32), but any sane limit will do because of
> the buf.buf_len test.
That would work, too. I don't know which way looks nicer.
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2025-05-02 12:41 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] < <CAG48ez1DPWhT2dhd1iptFawWjteh_=pZ4M6Yq5KKCq2DTArnqw@mail.gmail.com>
2025-05-02 12:41 ` [PATCH] tee: Prevent size calculation wraparound on 32-bit kernels David Laight
[not found] < <CAK8z29XWCujUdLXcHz075+xcix8HV2Mp3EtxxX9GB7vGjwi3HA@mail.gmail.com>
2025-04-30 13:00 ` Jens Wiklander
[not found] < <CAHUa44E_JZdYnGrReP0zWCP1wdu2BdJ9DSZZ3a2OiobRj61ThQ@mail.gmail.com>
2025-04-30 12:12 ` Rouven Czerwinski
2025-04-28 13:06 Jann Horn
2025-04-30 11:52 ` Jens Wiklander
2025-05-01 20:01 ` David Laight
2025-05-02 12:28 ` Jann Horn
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox