drivers/perf: arm_spe: Don't error on high-order pages for aux buf

Since commit 5768402fd9 ("perf/ring_buffer: Use high order allocations
for AUX buffers optimistically"), the perf core tends to back aux buffer
allocations with high-order pages with the order encoded in the
PagePrivate data. The Arm SPE driver explicitly rejects such pages,
causing the perf tool to fail with:

  | failed to mmap with 12 (Cannot allocate memory)

In actual fact, we can simply treat these pages just like any other
since the perf core takes care to populate the page array appropriately.
In theory we could try to map with PMDs where possible, but for now,
let's just get things working again.

Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Fixes: 5768402fd9 ("perf/ring_buffer: Use high order allocations for AUX buffers optimistically")
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Hanjun Guo <guohanjun@huawei.com>
Tested-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
This commit is contained in:
Will Deacon 2019-05-13 17:53:03 +01:00
parent a98d9ae937
commit 14ae42a6f0

View File

@ -855,16 +855,8 @@ static void *arm_spe_pmu_setup_aux(struct perf_event *event, void **pages,
if (!pglist)
goto out_free_buf;
for (i = 0; i < nr_pages; ++i) {
struct page *page = virt_to_page(pages[i]);
if (PagePrivate(page)) {
pr_warn("unexpected high-order page for auxbuf!");
goto out_free_pglist;
}
for (i = 0; i < nr_pages; ++i)
pglist[i] = virt_to_page(pages[i]);
}
buf->base = vmap(pglist, nr_pages, VM_MAP, PAGE_KERNEL);
if (!buf->base)