[RESEND,v1,2/2] misc: fastrpc: detect privileged processes based on group ID
Commit Message
Get the information on privileged group IDs during rpmsg probing based
on DT property. Check if the process requesting an offload to remote
subsystem is privileged by comparing it's group ID with privileged
group ID. Initialization process attributes are updated for a
privileged process which is sent to remote process for resource
management.
Signed-off-by: Ekansh Gupta <quic_ekangupt@quicinc.com>
---
drivers/misc/fastrpc.c | 124 +++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 124 insertions(+)
Comments
On 07/06/2023 18:30, Ekansh Gupta wrote:
> Get the information on privileged group IDs during rpmsg probing based
> on DT property. Check if the process requesting an offload to remote
> subsystem is privileged by comparing it's group ID with privileged
> group ID. Initialization process attributes are updated for a
> privileged process which is sent to remote process for resource
> management.
>
> +
> static const struct file_operations fastrpc_fops = {
> .open = fastrpc_device_open,
> .release = fastrpc_device_release,
> @@ -2277,6 +2396,10 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
> if (!data)
> return -ENOMEM;
>
> + err = fastrpc_init_privileged_gids(rdev, "qcom,fastrpc-gids", &data->gidlist);
> + if (err)
> + dev_err(rdev, "Privileged gids init failed.\n");
> +
What about error paths? No need for cleanup?
Best regards,
Krzysztof
On 6/8/2023 12:17 AM, Krzysztof Kozlowski wrote:
> On 07/06/2023 18:30, Ekansh Gupta wrote:
>> Get the information on privileged group IDs during rpmsg probing based
>> on DT property. Check if the process requesting an offload to remote
>> subsystem is privileged by comparing it's group ID with privileged
>> group ID. Initialization process attributes are updated for a
>> privileged process which is sent to remote process for resource
>> management.
>>
>
>
>
>> +
>> static const struct file_operations fastrpc_fops = {
>> .open = fastrpc_device_open,
>> .release = fastrpc_device_release,
>> @@ -2277,6 +2396,10 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
>> if (!data)
>> return -ENOMEM;
>>
>> + err = fastrpc_init_privileged_gids(rdev, "qcom,fastrpc-gids", &data->gidlist);
>> + if (err)
>> + dev_err(rdev, "Privileged gids init failed.\n");
>> +
>
> What about error paths? No need for cleanup?
>
All the necessary clean-up is added as part of
fastrpc_init_privileged_gids error path. There is no requirement to have
any additional handling in error path other that error log. Also there
is no intention to fail the probe in case gid information is not
properly read.
Thanks,
Ekansh
> Best regards,
> Krzysztof
>
On 08/06/2023 12:44, Ekansh Gupta wrote:
>
>
> On 6/8/2023 12:17 AM, Krzysztof Kozlowski wrote:
>> On 07/06/2023 18:30, Ekansh Gupta wrote:
>>> Get the information on privileged group IDs during rpmsg probing based
>>> on DT property. Check if the process requesting an offload to remote
>>> subsystem is privileged by comparing it's group ID with privileged
>>> group ID. Initialization process attributes are updated for a
>>> privileged process which is sent to remote process for resource
>>> management.
>>>
>>
>>
>>
>>> +
>>> static const struct file_operations fastrpc_fops = {
>>> .open = fastrpc_device_open,
>>> .release = fastrpc_device_release,
>>> @@ -2277,6 +2396,10 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
>>> if (!data)
>>> return -ENOMEM;
>>>
>>> + err = fastrpc_init_privileged_gids(rdev, "qcom,fastrpc-gids", &data->gidlist);
>>> + if (err)
>>> + dev_err(rdev, "Privileged gids init failed.\n");
>>> +
>>
>> What about error paths? No need for cleanup?
>>
> All the necessary clean-up is added as part of
> fastrpc_init_privileged_gids error path. There is no requirement to have
Where? How that code is called after fastrpc_device_register() failure?
Or after of_platform_populate() failure?
Please show me the code flow.
> any additional handling in error path other that error log. Also there
> is no intention to fail the probe in case gid information is not
> properly read.
This is not related. I don't talk about fastrpc_init_privileged_gids()
failures. Look where did I leave my comment.
Review comments are placed in proper places, not in unrelated parts of
code. The placement of review comment is important as this is the
context of bug in your patch.
Best regards,
Krzysztof
@@ -21,6 +21,7 @@
#include <linux/firmware/qcom/qcom_scm.h>
#include <uapi/misc/fastrpc.h>
#include <linux/of_reserved_mem.h>
+#include <linux/cred.h>
#define ADSP_DOMAIN_ID (0)
#define MDSP_DOMAIN_ID (1)
@@ -166,6 +167,11 @@ struct fastrpc_mem_unmap_req_msg {
u64 len;
};
+struct gid_list {
+ u32 *gids;
+ u32 gidcount;
+};
+
struct fastrpc_msg {
int pid; /* process group id */
int tid; /* thread id */
@@ -277,6 +283,7 @@ struct fastrpc_channel_ctx {
struct fastrpc_device *fdevice;
struct fastrpc_buf *remote_heap;
struct list_head invoke_interrupted_mmaps;
+ struct gid_list gidlist;
bool secure;
bool unsigned_support;
u64 dma_mask;
@@ -305,6 +312,7 @@ struct fastrpc_user {
spinlock_t lock;
/* lock for allocations */
struct mutex mutex;
+ struct gid_list gidlist;
};
static void fastrpc_free_map(struct kref *ref)
@@ -522,6 +530,31 @@ static void fastrpc_context_put_wq(struct work_struct *work)
}
#define CMP(aa, bb) ((aa) == (bb) ? 0 : (aa) < (bb) ? -1 : 1)
+
+static u32 sorted_lists_intersection(u32 *listA,
+ u32 lenA, u32 *listB, u32 lenB)
+{
+ u32 i = 0, j = 0;
+
+ while (i < lenA && j < lenB) {
+ if (listA[i] < listB[j])
+ i++;
+ else if (listA[i] > listB[j])
+ j++;
+ else
+ return listA[i];
+ }
+ return 0;
+}
+
+static int uint_cmp_func(const void *p1, const void *p2)
+{
+ u32 a1 = *((u32 *)p1);
+ u32 a2 = *((u32 *)p2);
+
+ return CMP(a1, a2);
+}
+
static int olaps_cmp(const void *a, const void *b)
{
struct fastrpc_buf_overlap *pa = (struct fastrpc_buf_overlap *)a;
@@ -1230,6 +1263,50 @@ static bool is_session_rejected(struct fastrpc_user *fl, bool unsigned_pd_reques
return false;
}
+static int fastrpc_get_process_gids(struct gid_list *gidlist)
+{
+ struct group_info *group_info = get_current_groups();
+ int i, num_gids;
+ u32 *gids = NULL;
+
+ if (!group_info)
+ return -EFAULT;
+
+ num_gids = group_info->ngroups + 1;
+ gids = kcalloc(num_gids, sizeof(u32), GFP_KERNEL);
+ if (!gids)
+ return -ENOMEM;
+
+ /* Get the real GID */
+ gids[0] = __kgid_val(current_gid());
+
+ /* Get the supplemental GIDs */
+ for (i = 1; i < num_gids; i++)
+ gids[i] = __kgid_val(group_info->gid[i - 1]);
+
+ sort(gids, num_gids, sizeof(*gids), uint_cmp_func, NULL);
+ gidlist->gids = gids;
+ gidlist->gidcount = num_gids;
+
+ return 0;
+}
+
+static void fastrpc_check_privileged_process(struct fastrpc_user *fl,
+ struct fastrpc_init_create *init)
+{
+ u32 gid = sorted_lists_intersection(fl->gidlist.gids,
+ fl->gidlist.gidcount, fl->cctx->gidlist.gids,
+ fl->cctx->gidlist.gidcount);
+
+ /* disregard any privilege bits from userspace */
+ init->attrs &= (~FASTRPC_MODE_PRIVILEGED);
+ if (gid) {
+ dev_info(&fl->cctx->rpdev->dev, "%s: %s (PID %d, GID %u) is a privileged process\n",
+ __func__, current->comm, fl->tgid, gid);
+ init->attrs |= FASTRPC_MODE_PRIVILEGED;
+ }
+}
+
static int fastrpc_init_create_static_process(struct fastrpc_user *fl,
char __user *argp)
{
@@ -1386,6 +1463,8 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl,
goto err;
}
+ fastrpc_get_process_gids(&fl->gidlist);
+
inbuf.pgid = fl->tgid;
inbuf.namelen = strlen(current->comm) + 1;
inbuf.filelen = init.filelen;
@@ -1400,6 +1479,8 @@ static int fastrpc_init_create_process(struct fastrpc_user *fl,
goto err;
}
+ fastrpc_check_privileged_process(fl, &init);
+
memlen = ALIGN(max(INIT_FILELEN_MAX, (int)init.filelen * 4),
1024 * 1024);
err = fastrpc_buf_alloc(fl, fl->sctx->dev, memlen,
@@ -1519,6 +1600,7 @@ static int fastrpc_device_release(struct inode *inode, struct file *file)
spin_lock_irqsave(&cctx->lock, flags);
list_del(&fl->user);
spin_unlock_irqrestore(&cctx->lock, flags);
+ kfree(fl->gidlist.gids);
if (fl->init_mem)
fastrpc_buf_free(fl->init_mem);
@@ -2118,6 +2200,43 @@ static long fastrpc_device_ioctl(struct file *file, unsigned int cmd,
return err;
}
+static int fastrpc_init_privileged_gids(struct device *dev, char *prop_name,
+ struct gid_list *gidlist)
+{
+ int err = 0;
+ u32 len = 0, i;
+ u32 *gids = NULL;
+
+ if (!of_find_property(dev->of_node, prop_name, &len))
+ return 0;
+ if (len == 0)
+ return 0;
+
+ len /= sizeof(u32);
+ gids = kcalloc(len, sizeof(u32), GFP_KERNEL);
+ if (!gids)
+ return -ENOMEM;
+
+ for (i = 0; i < len; i++) {
+ err = of_property_read_u32_index(dev->of_node, prop_name,
+ i, &gids[i]);
+ if (err) {
+ dev_err(dev, "%s: failed to read GID %u\n",
+ __func__, i);
+ goto read_error;
+ }
+ dev_info(dev, "adsprpc: %s: privileged GID: %u\n", __func__, gids[i]);
+ }
+ sort(gids, len, sizeof(*gids), uint_cmp_func, NULL);
+ gidlist->gids = gids;
+ gidlist->gidcount = len;
+
+ return 0;
+read_error:
+ kfree(gids);
+ return err;
+}
+
static const struct file_operations fastrpc_fops = {
.open = fastrpc_device_open,
.release = fastrpc_device_release,
@@ -2277,6 +2396,10 @@ static int fastrpc_rpmsg_probe(struct rpmsg_device *rpdev)
if (!data)
return -ENOMEM;
+ err = fastrpc_init_privileged_gids(rdev, "qcom,fastrpc-gids", &data->gidlist);
+ if (err)
+ dev_err(rdev, "Privileged gids init failed.\n");
+
if (vmcount) {
data->vmcount = vmcount;
data->perms = BIT(QCOM_SCM_VMID_HLOS);
@@ -2382,6 +2505,7 @@ static void fastrpc_rpmsg_remove(struct rpmsg_device *rpdev)
if (cctx->remote_heap)
fastrpc_buf_free(cctx->remote_heap);
+ kfree(cctx->gidlist.gids);
of_platform_depopulate(&rpdev->dev);
fastrpc_channel_ctx_put(cctx);