Re: [PATCH v10 4/11] LSM: syscalls for current process attributes

[Casey Schaufler]

On 6/7/2023 3:32 PM, Paul Moore wrote:
> On Apr 28, 2023 Casey Schaufler <casey@xxxxxxxxxxxxxxxx> wrote:
>> Create a system call lsm_get_self_attr() to provide the security
>> module maintained attributes of the current process.
>> Create a system call lsm_set_self_attr() to set a security
>> module maintained attribute of the current process.
>> Historically these attributes have been exposed to user space via
>> entries in procfs under /proc/self/attr.
>>
>> The attribute value is provided in a lsm_ctx structure. The structure
>> identifies the size of the attribute, and the attribute value. The format
>> of the attribute value is defined by the security module. A flags field
>> is included for LSM specific information. It is currently unused and must
>> be 0. The total size of the data, including the lsm_ctx structure and any
>> padding, is maintained as well.
>>
>> struct lsm_ctx {
>>         __u64 id;
>>         __u64 flags;
>>         __u64 len;
>>         __u64 ctx_len;
>>         __u8 ctx[];
>> };
>>
>> Two new LSM hooks are used to interface with the LSMs.
>> security_getselfattr() collects the lsm_ctx values from the
>> LSMs that support the hook, accounting for space requirements.
>> security_setselfattr() identifies which LSM the attribute is
>> intended for and passes it along.
>>
>> Signed-off-by: Casey Schaufler <casey@xxxxxxxxxxxxxxxx>
>> Reviewed-by: Kees Cook <keescook@xxxxxxxxxxxx>
>> ---
>>  Documentation/userspace-api/lsm.rst |  15 ++++
>>  include/linux/lsm_hook_defs.h       |   4 +
>>  include/linux/lsm_hooks.h           |   9 +++
>>  include/linux/security.h            |  19 +++++
>>  include/linux/syscalls.h            |   5 ++
>>  include/uapi/linux/lsm.h            |  36 +++++++++
>>  kernel/sys_ni.c                     |   4 +
>>  security/Makefile                   |   1 +
>>  security/lsm_syscalls.c             |  55 ++++++++++++++
>>  security/security.c                 | 112 ++++++++++++++++++++++++++++
>>  10 files changed, 260 insertions(+)
>>  create mode 100644 security/lsm_syscalls.c
> ..
>
>> diff --git a/security/security.c b/security/security.c
>> index 5a48b1b539e5..94b78bfd06b9 100644
>> --- a/security/security.c
>> +++ b/security/security.c
>> @@ -2176,6 +2176,118 @@ void security_d_instantiate(struct dentry *dentry, struct inode *inode)
>>  }
>>  EXPORT_SYMBOL(security_d_instantiate);
>>  
>> +/**
>> + * security_getselfattr - Read an LSM attribute of the current process.
>> + * @attr: which attribute to return
>> + * @ctx: the user-space destination for the information, or NULL
>> + * @size: the size of space available to receive the data
>> + * @flags: special handling options. LSM_FLAG_SINGLE indicates that only
>> + * attributes associated with the LSM identified in the passed @ctx be
>> + * reported
>> + *
>> + * Returns the number of attributes found on success, negative value
>> + * on error. @size is reset to the total size of the data.
>> + * If @size is insufficient to contain the data -E2BIG is returned.
>> + */
>> +int security_getselfattr(unsigned int attr, struct lsm_ctx __user *ctx,
>> +			 size_t __user *size, u32 flags)
>> +{
>> +	struct security_hook_list *hp;
>> +	struct lsm_ctx lctx = { .id = LSM_ID_UNDEF, };
>> +	u8 __user *base = (u8 __user *)ctx;
>> +	size_t total = 0;
>> +	size_t entrysize;
>> +	size_t left;
>> +	bool toobig = false;
>> +	int count = 0;
>> +	int rc;
>> +
>> +	if (attr == 0)
>> +		return -EINVAL;
>> +	if (size == NULL)
>> +		return -EINVAL;
>> +	if (get_user(left, size))
>> +		return -EFAULT;
>> +
>> +	if ((flags & LSM_FLAG_SINGLE) == LSM_FLAG_SINGLE) {
>> +		if (!ctx)
>> +			return -EINVAL;
>> +		if (copy_struct_from_user(&lctx, sizeof(lctx), ctx, left))
>> +			return -EFAULT;
>> +		if (lctx.id == LSM_ID_UNDEF)
>> +			return -EINVAL;
>> +	} else if (flags) {
>> +		return -EINVAL;
>> +	}
>> +
>> +	hlist_for_each_entry(hp, &security_hook_heads.getselfattr, list) {
>> +		if (lctx.id != LSM_ID_UNDEF && lctx.id != hp->lsmid->id)
>> +			continue;
> I think we're missing a copy_struct_from_user() call somewhere; how
> does @lctx get populated in the non-LSM_FLAG_SINGLE case?  How does
> it move to the next entry in the buffer?

@lctx.id is initialized to LSM_ID_UNDEF in the declaration. In the
non-LSM_FLAG_SINGLE case the lsmctx.id != LSM_ID_UNDEF test will always
fail, thus avoiding the continue. It is only necessary to do the
copy_struct_from_user() call in the LSM_FLAG_SINGLE case.

> Am I missing something obvious?

Yes.

>   Was this code tested?

Yes. See patch 11/11.

>
>> +		entrysize = left;
>> +		if (base)
>> +			ctx = (struct lsm_ctx __user *)(base + total);
>> +		rc = hp->hook.getselfattr(attr, ctx, &entrysize, flags);
>> +		if (rc == -EOPNOTSUPP) {
>> +			rc = 0;
>> +			continue;
>> +		}
>> +		if (rc == -E2BIG) {
>> +			toobig = true;
>> +			left = 0;
>> +			continue;
>> +		}
>> +		if (rc < 0)
>> +			return rc;
>> +
>> +		left -= entrysize;
>> +		total += entrysize;
>> +		count += rc;
>> +	}
>> +	if (put_user(total, size))
>> +		return -EFAULT;
>> +	if (toobig)
>> +		return -E2BIG;
>> +	if (count == 0)
>> +		return LSM_RET_DEFAULT(getselfattr);
>> +	return count;
>> +}
>> +
>> +/**
>> + * security_setselfattr - Set an LSM attribute on the current process.
>> + * @attr: which attribute to set
>> + * @ctx: the user-space source for the information
>> + * @size: the size of the data
>> + * @flags: reserved for future use, must be 0
>> + *
>> + * Set an LSM attribute for the current process. The LSM, attribute
>> + * and new value are included in @ctx.
>> + *
>> + * Returns 0 on success, -EINVAL if the input is inconsistent, -EFAULT
>> + * if the user buffer is inaccessible or an LSM specific failure.
>> + */
>> +int security_setselfattr(unsigned int attr, struct lsm_ctx __user *ctx,
>> +			 size_t size, u32 flags)
>> +{
>> +	struct security_hook_list *hp;
>> +	struct lsm_ctx lctx;
>> +
>> +	if (flags)
>> +		return -EINVAL;
> Once again, I don't see a reasonable way to support setting an
> attribute across multiple LSMs, but for the sake of consistency
> across both the getselfattr and setselfattr syscalls, what do
> people think about *requiring* the LSM_FLAG_SINGLE flag here in
> the setselfattr syscalls since it can only operate on a single
> LSM at a time?  A zero flag value would return -EINVAL.

Requiring a pointless flag seems like an API no-no to me.

>
>> +	if (size < sizeof(*ctx))
>> +		return -EINVAL;
>> +	if (copy_from_user(&lctx, ctx, sizeof(*ctx)))
>> +		return -EFAULT;
>> +	if (size < lctx.len || size < lctx.ctx_len + sizeof(ctx) ||
>> +	    lctx.len < lctx.ctx_len + sizeof(ctx))
>> +		return -EINVAL;
>> +
>> +	hlist_for_each_entry(hp, &security_hook_heads.setselfattr, list)
>> +		if ((hp->lsmid->id) == lctx.id)
>> +			return hp->hook.setselfattr(attr, ctx, size, flags);
>> +
>> +	return LSM_RET_DEFAULT(setselfattr);
>> +}
>> +
>>  int security_getprocattr(struct task_struct *p, int lsmid, const char *name,
>>  			 char **value)
>>  {
>> -- 
>> 2.39.2
> --
> paul-moore.com