#ifndef CSUM_H
#define	CSUM_H

#include <netinet/in.h>
#include <netinet/ip.h>

#include "built_in.h"

static inline unsigned short csum(unsigned short *buf, int nwords)
{
	unsigned long sum;

	for (sum = 0; nwords > 0; nwords--)
		sum += *buf++;
	sum = (sum >> 16) + (sum & 0xffff);
	sum += (sum >> 16);

	return ~sum;
}

static inline uint16_t calc_csum(void *addr, size_t len,
				 int ccsum __maybe_unused)
{
	return csum(addr, len >> 1);
}

static inline uint16_t csum_expected(uint16_t sum, uint16_t computed_sum)
{
	uint32_t shouldbe;

	shouldbe = sum;
	shouldbe += ntohs(computed_sum);
	shouldbe = (shouldbe & 0xFFFF) + (shouldbe >> 16);
	shouldbe = (shouldbe & 0xFFFF) + (shouldbe >> 16);

	return shouldbe;
}

/* Taken and modified from tcpdump, Copyright belongs to them! */

struct cksum_vec {
	const uint8_t *ptr;
	int len;
};

#define ADDCARRY(x)		\
	do { if ((x) > 65535)	\
		(x) -= 65535;	\
	} while (0)

#define REDUCE						\
	do {						\
		l_util.l = sum;				\
		sum = l_util.s[0] + l_util.s[1];	\
		ADDCARRY(sum);				\
	} while (0)

static inline uint16_t __in_cksum(const struct cksum_vec *vec, int veclen)
{
	const uint16_t *w;
	int sum = 0, mlen = 0;
	int byte_swapped = 0;
	union {
		uint8_t c[2];
		uint16_t s;
	} s_util;
	union {
		uint16_t s[2];
		uint32_t l;
	} l_util;

	for (; veclen != 0; vec++, veclen--) {
		if (vec->len == 0)
			continue;

		w = (const uint16_t *) (void *) vec->ptr;

		if (mlen == -1) {
			s_util.c[1] = *(const uint8_t *) w;
			sum += s_util.s;
			w = (const uint16_t *) (void *) ((const uint8_t *) w + 1);
			mlen = vec->len - 1;
		} else
			mlen = vec->len;

		if ((1 & (unsigned long) w) && (mlen > 0)) {
			REDUCE;
			sum <<= 8;
			s_util.c[0] = *(const uint8_t *) w;
			w = (const uint16_t *) (void *) ((const uint8_t *) w + 1);
			mlen--;
			byte_swapped = 1;
		}

		while ((mlen -= 32) >= 0) {
			sum +=  w[0]; sum +=  w[1]; sum +=  w[2]; sum +=  w[3];
			sum +=  w[4]; sum +=  w[5]; sum +=  w[6]; sum +=  w[7];
			sum +=  w[8]; sum +=  w[9]; sum += w[10]; sum += w[11];
			sum += w[12]; sum += w[13]; sum += w[14]; sum += w[15];
			w += 16;
		}

		mlen += 32;

		while ((mlen -= 8) >= 0) {
			sum += w[0]; sum += w[1]; sum += w[2]; sum += w[3];
			w += 4;
		}

		mlen += 8;

		if (mlen == 0 && byte_swapped == 0)
			continue;

		REDUCE;

		while ((mlen -= 2) >= 0) {
			sum += *w++;
		}

		if (byte_swapped) {
			REDUCE;
			sum <<= 8;
			byte_swapped = 0;

			if (mlen == -1) {
				s_util.c[1] = *(const uint8_t *) w;
				sum += s_util.s;
				mlen = 0;
			} else
				mlen = -1;
		} else if (mlen == -1)
			s_util.c[0] = *(const uint8_t *) w;
	}

	if (mlen == -1) {
		s_util.c[1] = 0;
		sum += s_util.s;
	}

	REDUCE;

	return (~sum & 0xffff);
}

static inline uint16_t p4_csum(const struct ip *ip, const uint8_t *data,
			       uint16_t len, uint8_t next_proto)
{
	struct cksum_vec vec[2];
	struct pseudo_hdr {
		uint32_t src;
		uint32_t dst;
		uint8_t mbz;
		uint8_t proto;
		uint16_t len;
	} ph;

	memset(&ph, 0, sizeof(ph));
	ph.len = htons(len);
	ph.mbz = 0;
	ph.proto = next_proto;
	ph.src = ip->ip_src.s_addr;
	ph.dst = ip->ip_dst.s_addr;

	vec[0].ptr = (const uint8_t *) (void *) &ph;
	vec[0].len = sizeof(ph);

	vec[1].ptr = data;
	vec[1].len = len;

	return __in_cksum(vec, 2);
}

#endif /* CSUM_H */
'selected'>3</option><option value='4'>4</option><option value='5'>5</option><option value='6'>6</option><option value='7'>7</option><option value='8'>8</option><option value='9'>9</option><option value='10'>10</option><option value='15'>15</option><option value='20'>20</option><option value='25'>25</option><option value='30'>30</option><option value='35'>35</option><option value='40'>40</option></select></td></tr><tr><td class='label'>space:</td><td class='ctrl'><select name='ignorews' onchange='this.form.submit();'><option value='0' selected='selected'>include</option><option value='1'>ignore</option></select></td></tr><tr><td class='label'>mode:</td><td class='ctrl'><select name='dt' onchange='this.form.submit();'><option value='0' selected='selected'>unified</option><option value='1'>ssdiff</option><option value='2'>stat only</option></select></td></tr><tr><td/><td class='ctrl'><noscript><input type='submit' value='reload'/></noscript></td></tr></table></form></div><table summary='commit info' class='commit-info'>
<tr><th>author</th><td>Dave Airlie &lt;airlied@redhat.com&gt;</td><td class='right'>2016-10-24 15:27:59 +1000</td></tr>
<tr><th>committer</th><td>Dave Airlie &lt;airlied@redhat.com&gt;</td><td class='right'>2016-10-26 15:45:38 +1000</td></tr>
<tr><th>commit</th><td colspan='2' class='oid'><a href='/cgit.cgi/linux/net-next.git/commit/Documentation?id=8ef4227615e158faa4ee85a1d6466782f7e22f2f'>8ef4227615e158faa4ee85a1d6466782f7e22f2f</a> (<a href='/cgit.cgi/linux/net-next.git/patch/Documentation?id=8ef4227615e158faa4ee85a1d6466782f7e22f2f'>patch</a>)</td></tr>
<tr><th>tree</th><td colspan='2' class='oid'><a href='/cgit.cgi/linux/net-next.git/tree/?id=8ef4227615e158faa4ee85a1d6466782f7e22f2f'>b5f5ef11dd409790679bf53daf6ba3ed586082e2</a> /<a href='/cgit.cgi/linux/net-next.git/tree/Documentation?id=8ef4227615e158faa4ee85a1d6466782f7e22f2f'>Documentation</a></td></tr>
<tr><th>parent</th><td colspan='2' class='oid'><a href='/cgit.cgi/linux/net-next.git/commit/Documentation?id=07d9a380680d1c0eb51ef87ff2eab5c994949e69'>07d9a380680d1c0eb51ef87ff2eab5c994949e69</a> (<a href='/cgit.cgi/linux/net-next.git/diff/Documentation?id=8ef4227615e158faa4ee85a1d6466782f7e22f2f&amp;id2=07d9a380680d1c0eb51ef87ff2eab5c994949e69'>diff</a>)</td></tr></table>
<div class='commit-subject'>x86/io: add interface to reserve io memtype for a resource range. (v1.1)</div><div class='commit-msg'>A recent change to the mm code in:
87744ab3832b mm: fix cache mode tracking in vm_insert_mixed()

started enforcing checking the memory type against the registered list for
amixed pfn insertion mappings. It happens that the drm drivers for a number
of gpus relied on this being broken. Currently the driver only inserted
VRAM mappings into the tracking table when they came from the kernel,
and userspace mappings never landed in the table. This led to a regression
where all the mapping end up as UC instead of WC now.

I've considered a number of solutions but since this needs to be fixed
in fixes and not next, and some of the solutions were going to introduce
overhead that hadn't been there before I didn't consider them viable at
this stage. These mainly concerned hooking into the TTM io reserve APIs,
but these API have a bunch of fast paths I didn't want to unwind to add
this to.

The solution I've decided on is to add a new API like the arch_phys_wc
APIs (these would have worked but wc_del didn't take a range), and
use them from the drivers to add a WC compatible mapping to the table
for all VRAM on those GPUs. This means we can then create userspace
mapping that won't get degraded to UC.

v1.1: use CONFIG_X86_PAT + add some comments in io.h

Cc: Toshi Kani &lt;toshi.kani@hp.com&gt;
Cc: Borislav Petkov &lt;bp@alien8.de&gt;
Cc: H. Peter Anvin &lt;hpa@zytor.com&gt;
Cc: Andy Lutomirski &lt;luto@kernel.org&gt;
Cc: Denys Vlasenko &lt;dvlasenk@redhat.com&gt;
Cc: Brian Gerst &lt;brgerst@gmail.com&gt;
Cc: x86@kernel.org
Cc: mcgrof@suse.com
Cc: Dan Williams &lt;dan.j.williams@intel.com&gt;
Acked-by: Ingo Molnar &lt;mingo@kernel.org&gt;
Reviewed-by: Thomas Gleixner &lt;tglx@linutronix.de&gt;
Signed-off-by: Dave Airlie &lt;airlied@redhat.com&gt;
</div><div class='diffstat-header'><a href='/cgit.cgi/linux/net-next.git/diff/?id=8ef4227615e158faa4ee85a1d6466782f7e22f2f'>Diffstat</a> (limited to 'Documentation')</div><table summary='diffstat' class='diffstat'>