| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Thu, 26 Oct 2023 10:44:28 -0700
From: Kees Cook <keescook@...omium.org>
To: Justin Stitt <justinstitt@...gle.com>
Cc: Don Brace <don.brace@...rochip.com>,
"James E.J. Bottomley" <jejb@...ux.ibm.com>,
"Martin K. Petersen" <martin.petersen@...cle.com>,
storagedev@...rochip.com, linux-scsi@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-hardening@...r.kernel.org
Subject: Re: [PATCH] scsi: hpsa: replace deprecated strncpy with
strscpy/kmemdup_nul
On Thu, Oct 26, 2023 at 01:47:32AM +0000, Justin Stitt wrote:
> strncpy() is deprecated for use on NUL-terminated destination strings
> [1] and as such we should prefer more robust and less ambiguous string
> interfaces.
>
> This whole process of 1) determining smaller length so we don't overread
> the buffer and 2) manually NUL-terminating our buffer so we can use in
> string APIs is handled implicitly by strscpy().
>
> Therefore, a suitable replacement is `strscpy` [2] due to the fact that
> it guarantees NUL-termination on the destination buffer without
> unnecessarily NUL-padding.
>
> For the last two strncpy() use cases in init_driver_version(), we can
> actually drop this function entirely.
>
> Firstly, we are kmalloc()'ing driver_version. Then, we are calling
> init_driver_version() which memset's it to 0 followed by a strncpy().
> This pattern of 1) allocating memory for a string, 2) setting all bytes
> to NUL, 3) copy bytes from another string + ensure NUL-padded
> destination is just an open-coded kmemdup_nul().
>
> The last case involves swapping kmalloc_array() for kcalloc() to give us
> a zero-filled two-element array for both old_driver_version and
> driver_version without needing the memset from init_driver_version().
>
> Now this code is easier to read and less fragile (no more ... - 1's) or
> min length checks and now we have guaranteed NUL-termination everywhere!
>
> Although perhaps there should be a macro for:
>
> Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
> Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2]
> Link: https://github.com/KSPP/linux/issues/90
> Cc: linux-hardening@...r.kernel.org
> Cc: Kees Cook <keescook@...omium.org>
> Signed-off-by: Justin Stitt <justinstitt@...gle.com>
> ---
> Note: build-tested only.
>
> Found with: $ rg "strncpy\("
> ---
> drivers/scsi/hpsa.c | 29 +++++++++++------------------
> 1 file changed, 11 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/scsi/hpsa.c b/drivers/scsi/hpsa.c
> index af18d20f3079..3376d4614fe5 100644
> --- a/drivers/scsi/hpsa.c
> +++ b/drivers/scsi/hpsa.c
> @@ -452,16 +452,15 @@ static ssize_t host_store_hp_ssd_smart_path_status(struct device *dev,
> struct device_attribute *attr,
> const char *buf, size_t count)
> {
> - int status, len;
> + int status;
> struct ctlr_info *h;
> struct Scsi_Host *shost = class_to_shost(dev);
> char tmpbuf[10];
>
> if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
> return -EACCES;
> - len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count;
> - strncpy(tmpbuf, buf, len);
> - tmpbuf[len] = '\0';
> + strscpy(tmpbuf, buf, count);
This is wrong -- "count" isn't the size of tmpbuf -- it's the size of
the source, i.e. strlen(buf).
> +
> if (sscanf(tmpbuf, "%d", &status) != 1)
> return -EINVAL;
And this is immediately using the tmpbuf for getting an int. All of this
should be replaced by kstrtoint().
> h = shost_to_hba(shost);
> @@ -476,16 +475,15 @@ static ssize_t host_store_raid_offload_debug(struct device *dev,
> struct device_attribute *attr,
> const char *buf, size_t count)
> {
> - int debug_level, len;
> + int debug_level;
> struct ctlr_info *h;
> struct Scsi_Host *shost = class_to_shost(dev);
> char tmpbuf[10];
>
> if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO))
> return -EACCES;
> - len = count > sizeof(tmpbuf) - 1 ? sizeof(tmpbuf) - 1 : count;
> - strncpy(tmpbuf, buf, len);
> - tmpbuf[len] = '\0';
> + strscpy(tmpbuf, buf, count);
> +
> if (sscanf(tmpbuf, "%d", &debug_level) != 1)
> return -EINVAL;
Same thing here.
> if (debug_level < 0)
> @@ -7234,24 +7232,19 @@ static int hpsa_controller_hard_reset(struct pci_dev *pdev,
> return 0;
> }
>
> -static void init_driver_version(char *driver_version, int len)
> -{
> - memset(driver_version, 0, len);
> - strncpy(driver_version, HPSA " " HPSA_DRIVER_VERSION, len - 1);
> -}
> -
> static int write_driver_ver_to_cfgtable(struct CfgTable __iomem *cfgtable)
> {
> char *driver_version;
> int i, size = sizeof(cfgtable->driver_version);
>
> - driver_version = kmalloc(size, GFP_KERNEL);
> + driver_version = kmemdup_nul(HPSA " " HPSA_DRIVER_VERSION, size,
> + GFP_KERNEL);
"size" isn't the length of the string here, so this results in an
over-read from the .data segment:
drivers/scsi/hpsa.c:#define HPSA "hpsa"
drivers/scsi/hpsa.c:#define HPSA_DRIVER_VERSION "3.4.20-200"
strlen(HSPA " " HPSA_DRIVER_VERSION) == 15 (16 with %NUL terminator)
sizeof(cfgtable->driver_version) == 32:
struct CfgTable {
...
u8 driver_version[32];
> if (!driver_version)
> return -ENOMEM;
>
> - init_driver_version(driver_version, size);
> for (i = 0; i < size; i++)
> writeb(driver_version[i], &cfgtable->driver_version[i]);
And then this will write garbage out to the driver for the 16 bytes
following the string... :(
Also, this thing is doing an alloc/free for a tiny string. That can just
be on the stack:
char driver_version[sizeof(cfgtable->driver_version)] = HPSA " " HPSA_DRIVER_VERSION;
No alloc/free, no strscpy, easy easy. (Since the string is explicitly
sized, the remaining space will be zero-initialized.)
> +
> kfree(driver_version);
> return 0;
> }
> @@ -7271,7 +7264,7 @@ static int controller_reset_failed(struct CfgTable __iomem *cfgtable)
> char *driver_ver, *old_driver_ver;
> int rc, size = sizeof(cfgtable->driver_version);
>
> - old_driver_ver = kmalloc_array(2, size, GFP_KERNEL);
> + old_driver_ver = kcalloc(2, size, GFP_KERNEL);
> if (!old_driver_ver)
> return -ENOMEM;
> driver_ver = old_driver_ver + size;
> @@ -7279,7 +7272,7 @@ static int controller_reset_failed(struct CfgTable __iomem *cfgtable)
> /* After a reset, the 32 bytes of "driver version" in the cfgtable
> * should have been changed, otherwise we know the reset failed.
> */
> - init_driver_version(old_driver_ver, size);
> + strscpy(old_driver_ver, HPSA " " HPSA_DRIVER_VERSION, size);
> read_driver_ver_from_cfgtable(cfgtable, driver_ver);
> rc = !memcmp(driver_ver, old_driver_ver, size);
> kfree(old_driver_ver);
This function is also wild -- it's allocating 2 strings (but at the same
time, and using offsets to get to them), and again -- why? Just use the
stack for 64 bytes:
char driver_ver[sizeof(cfgtable->driver_version)] = "";
char old_driver_ver[sizeof(cfgtable->driver_version)] = HPSA " " HPSA_DRIVER_VERSION;
-Kees
--
Kees Cook
Powered by blists - more mailing lists