Messages in this thread | | | From | Jens Wiklander <> | Date | Wed, 16 Feb 2022 08:44:39 +0100 | Subject | Re: [PATCH] tee: make tee_shm_register_kernel_buf vmalloc supported |
| |
On Sun, Feb 13, 2022 at 5:21 PM Phil Chang <phil.chang@mediatek.com> wrote: > > In some low-memory devices, it's hard to aquire large-orders pages, > this patch allowed user using scatter pages to register shm. > > Signed-off-by: Phil Chang <phil.chang@mediatek.com> > --- > drivers/tee/optee/call.c | 2 +- > drivers/tee/tee_shm.c | 40 ++++++++++++++++++++++++++++------------ > 2 files changed, 29 insertions(+), 13 deletions(-) > > diff --git a/drivers/tee/optee/call.c b/drivers/tee/optee/call.c > index bd49ec934060..2082e632adff 100644 > --- a/drivers/tee/optee/call.c > +++ b/drivers/tee/optee/call.c > @@ -362,7 +362,7 @@ int optee_check_mem_type(unsigned long start, size_t num_pages) > * Allow kernel address to register with OP-TEE as kernel > * pages are configured as normal memory only. > */ > - if (virt_addr_valid(start)) > + if (virt_addr_valid(start) || is_vmalloc_addr((void *)start)) > return 0; > > mmap_read_lock(mm); > diff --git a/drivers/tee/tee_shm.c b/drivers/tee/tee_shm.c > index f31e29e8f1ca..861403588485 100644 > --- a/drivers/tee/tee_shm.c > +++ b/drivers/tee/tee_shm.c > @@ -23,22 +23,38 @@ static void shm_put_kernel_pages(struct page **pages, size_t page_count) > static int shm_get_kernel_pages(unsigned long start, size_t page_count, > struct page **pages) > { > - struct kvec *kiov; > - size_t n; > int rc; > > - kiov = kcalloc(page_count, sizeof(*kiov), GFP_KERNEL); > - if (!kiov) > - return -ENOMEM; > - > - for (n = 0; n < page_count; n++) { > - kiov[n].iov_base = (void *)(start + n * PAGE_SIZE); > - kiov[n].iov_len = PAGE_SIZE; > + if (is_vmalloc_addr((void *)start)) { > + struct page *page; > + int i; > + > + for (i = 0; i < page_count; i++) {
page_count is of an unsigned type and i is of a signed type, this is bound to cause a warning. Please move the "n" variable below to the top of this function and use that as an index variable here too.
> + page = vmalloc_to_page((void *)(start + PAGE_SIZE * i)); > + if (!page) > + return -ENOMEM; > + > + get_page(page); > + pages[i] = page; > + } > + rc = page_count; > + } else { > + struct kvec *kiov; > + size_t n; > + > + kiov = kcalloc(page_count, sizeof(*kiov), GFP_KERNEL); > + if (!kiov) > + return -ENOMEM; > + > + for (n = 0; n < page_count; n++) { > + kiov[n].iov_base = (void *)(start + n * PAGE_SIZE); > + kiov[n].iov_len = PAGE_SIZE; > + } > + > + rc = get_kernel_pages(kiov, page_count, 0, pages); > + kfree(kiov); > } > > - rc = get_kernel_pages(kiov, page_count, 0, pages); > - kfree(kiov); > - > return rc; > } > > -- > 2.25.1
This patch makes sense. Please fix the comment above and post a v3 (with a changelog this time).
Cheers, Jens
| |