On 10/11/21 01:34, Yanling Song wrote:
get_unaligned_be*() is an inline which has the same function as our
current code and there is no difference on performance.
But current code is better when supporting old kernels since it is
implemented on SCSI spec and there is no dependency on
get_unaligned_be*(), which means the code can work on any version
of kernel.
So I prefer to keep current implementation. What's your opinion?
Hi Yanling,
On all architectures I'm familiar with get_unaligned_be*() is faster
than fetching individual bytes and combining these with shift
operations. As an example, x86 CPU's support unaligned memory accesses
and for these CPU's the Linux kernel translates get_unaligned_be*() into
a single (potentially unaligned) memory access.
Kernel drivers that are submitted for upstream inclusion should use the
upstream kernel API. Whether or not get_unaligned_be*() is available in
older kernels does not matter - since it is available in the upstream
kernel and since it makes code faster and easier to read, using that
function is strongly recommended. Additionally, it can happen that after
a driver has been accepted upstream that someone writes a Coccinelle
script to change all open-coded get_unaligned_be*() calls into
get_unaligned_be*() calls. Compatibility with older kernels would not be
accepted as a valid argument against such a patch.
I think that get_unaligned_be*() functions are supported since kernel
version 2.6.26, released in 2008, 13 years ago. Does this address your
concern about supporting older kernel versions?
Regarding supporting older kernel versions, a common approach is to
develop against the latest upstream kernel. To support older kernels,
include a header file called e.g. backport.h and in that header file
implement the latest kernel API in terms of older kernel functions. An
example:
#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0) && \
!defined(CONFIG_SUSE_KERNEL)
static inline void bio_set_dev(struct bio *bio,
struct block_device *bdev)
{
bio->bi_bdev = bdev;
}
#endif
Bart.
