Hello Shamir Rabinovitch,
The patch ff23dfa13457: ("IB: Pass only ib_udata in function prototypes")
leads to the following static checker warning:
drivers/infiniband/sw/rdmavt/cq.c:273 rvt_create_cq()
error: 'cq->ip' dereferencing possible ERR_PTR()
drivers/infiniband/sw/rdmavt/cq.c
217 /*
218 * Return the address of the WC as the offset to mmap.
219 * See rvt_mmap() for details.
220 */
221 if (udata && udata->outlen >= sizeof(__u64)) {
222 cq->ip = rvt_create_mmap_info(rdi, sz, udata, wc);
This warning is a false positive which is kind of involved to explain.
This function returns a mix of error pointers and NULL. Smatch parses
correctly so it knows that "cq->ip" isn't an error pointer here. But
really it shouldn't return error pointers.
When a function returns error pointers and NULL the NULL return should
be treated as a special success case. For example, we could say
"p = get_some_feature();" If it get_some_feature() returns an error
pointer that means kmalloc() failed or whatever, but if it returns a
NULL that means the feature is disabled.
223 if (!cq->ip) {
224 err = -ENOMEM;
225 goto bail_wc;
226 }
227
228 err = ib_copy_to_udata(udata, &cq->ip->offset,
229 sizeof(cq->ip->offset));
230 if (err)
231 goto bail_ip;
232 }
regards,
dan carpenter
