diff --git a/iscsi-scst/kernel/patches/rhel/put_page_callback-3.10.0-123.patch b/iscsi-scst/kernel/patches/rhel/put_page_callback-3.10.0-123.patch new file mode 100644 index 000000000..534c659be --- /dev/null +++ b/iscsi-scst/kernel/patches/rhel/put_page_callback-3.10.0-123.patch @@ -0,0 +1,399 @@ +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/block/drbd/drbd_receiver.c ./drivers/block/drbd/drbd_receiver.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/block/drbd/drbd_receiver.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./drivers/block/drbd/drbd_receiver.c 2014-08-07 09:14:18.606676153 +0200 +@@ -130,7 +130,7 @@ static int page_chain_free(struct page * + struct page *tmp; + int i = 0; + page_chain_for_each_safe(page, tmp) { +- put_page(page); ++ net_put_page(page); + ++i; + } + return i; +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/macvtap.c ./drivers/net/macvtap.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/macvtap.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./drivers/net/macvtap.c 2014-08-07 09:14:18.606676153 +0200 +@@ -597,7 +597,7 @@ static int zerocopy_sg_from_iovec(struct + int j; + + for (j = 0; j < num_pages; j++) +- put_page(page[i + j]); ++ net_put_page(page[i + j]); + return -EFAULT; + } + truesize = size * PAGE_SIZE; +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/tun.c ./drivers/net/tun.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/tun.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./drivers/net/tun.c 2014-08-07 09:14:18.606676153 +0200 +@@ -1015,7 +1015,7 @@ static int zerocopy_sg_from_iovec(struct + int j; + + for (j = 0; j < num_pages; j++) +- put_page(page[i + j]); ++ net_put_page(page[i + j]); + return -EFAULT; + } + truesize = size * PAGE_SIZE; +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/vmxnet3/vmxnet3_drv.c ./drivers/net/vmxnet3/vmxnet3_drv.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/vmxnet3/vmxnet3_drv.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./drivers/net/vmxnet3/vmxnet3_drv.c 2014-08-07 09:14:18.606676153 +0200 +@@ -1360,7 +1360,7 @@ vmxnet3_rq_cleanup(struct vmxnet3_rx_que + rq->buf_info[ring_idx][i].page) { + pci_unmap_page(adapter->pdev, rxd->addr, + rxd->len, PCI_DMA_FROMDEVICE); +- put_page(rq->buf_info[ring_idx][i].page); ++ net_put_page(rq->buf_info[ring_idx][i].page); + rq->buf_info[ring_idx][i].page = NULL; + } + } +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/xen-netback/netback.c ./drivers/net/xen-netback/netback.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/drivers/net/xen-netback/netback.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./drivers/net/xen-netback/netback.c 2014-08-07 09:14:18.606676153 +0200 +@@ -1292,7 +1292,7 @@ static void xen_netbk_fill_frags(struct + skb->truesize += txp->size; + + /* Take an extra reference to offset xen_netbk_idx_release */ +- get_page(netbk->mmap_pages[pending_idx]); ++ net_get_page(netbk->mmap_pages[pending_idx]); + xen_netbk_idx_release(netbk, pending_idx, XEN_NETIF_RSP_OKAY); + } + } +@@ -1774,7 +1774,7 @@ static void xen_netbk_idx_release(struct + } while (!pending_tx_is_head(netbk, peek)); + + netbk->mmap_pages[pending_idx]->mapping = 0; +- put_page(netbk->mmap_pages[pending_idx]); ++ net_put_page(netbk->mmap_pages[pending_idx]); + netbk->mmap_pages[pending_idx] = NULL; + } + +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/mm_types.h ./include/linux/mm_types.h +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/mm_types.h 2014-07-16 20:25:31.000000000 +0200 ++++ ./include/linux/mm_types.h 2014-08-07 09:14:18.606676153 +0200 +@@ -186,6 +186,17 @@ struct page { + #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS + int _last_cpupid; + #endif ++ ++#if defined(CONFIG_TCP_ZERO_COPY_TRANSFER_COMPLETION_NOTIFICATION) ++ /* ++ * Used to implement support for notification on zero-copy TCP transfer ++ * completion. It might look as not good to have this field here and ++ * it's better to have it in struct sk_buff, but it would make the code ++ * much more complicated and fragile, since all skb then would have to ++ * contain only pages with the same value in this field. ++ */ ++ void *net_priv; ++#endif + } + /* + * The struct page can be forced to be double word aligned so that atomic ops +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/net.h ./include/linux/net.h +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/net.h 2014-07-16 20:25:31.000000000 +0200 ++++ ./include/linux/net.h 2014-08-07 09:14:18.606676153 +0200 +@@ -19,6 +19,7 @@ + #define _LINUX_NET_H + + #include ++#include + #include + #include + #include /* For O_CLOEXEC and O_NONBLOCK */ +@@ -278,6 +279,45 @@ extern int kernel_sock_ioctl(struct sock + extern int kernel_sock_shutdown(struct socket *sock, + enum sock_shutdown_cmd how); + ++#if defined(CONFIG_TCP_ZERO_COPY_TRANSFER_COMPLETION_NOTIFICATION) ++/* Support for notification on zero-copy TCP transfer completion */ ++typedef void (*net_get_page_callback_t)(struct page *page); ++typedef void (*net_put_page_callback_t)(struct page *page); ++ ++extern net_get_page_callback_t net_get_page_callback; ++extern net_put_page_callback_t net_put_page_callback; ++ ++extern int net_set_get_put_page_callbacks( ++ net_get_page_callback_t get_callback, ++ net_put_page_callback_t put_callback); ++ ++/* ++ * See comment for net_set_get_put_page_callbacks() why those functions ++ * don't need any protection. ++ */ ++static inline void net_get_page(struct page *page) ++{ ++ if (page->net_priv != 0) ++ net_get_page_callback(page); ++ get_page(page); ++} ++static inline void net_put_page(struct page *page) ++{ ++ if (page->net_priv != 0) ++ net_put_page_callback(page); ++ put_page(page); ++} ++#else ++static inline void net_get_page(struct page *page) ++{ ++ get_page(page); ++} ++static inline void net_put_page(struct page *page) ++{ ++ put_page(page); ++} ++#endif /* CONFIG_TCP_ZERO_COPY_TRANSFER_COMPLETION_NOTIFICATION */ ++ + #define MODULE_ALIAS_NETPROTO(proto) \ + MODULE_ALIAS("net-pf-" __stringify(proto)) + +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/skbuff.h ./include/linux/skbuff.h +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/skbuff.h 2014-07-16 20:25:31.000000000 +0200 ++++ ./include/linux/skbuff.h 2014-08-07 09:14:18.606676153 +0200 +@@ -2036,7 +2036,7 @@ static inline struct page *skb_frag_page + */ + static inline void __skb_frag_ref(skb_frag_t *frag) + { +- get_page(skb_frag_page(frag)); ++ net_get_page(skb_frag_page(frag)); + } + + /** +@@ -2059,7 +2059,7 @@ static inline void skb_frag_ref(struct s + */ + static inline void __skb_frag_unref(skb_frag_t *frag) + { +- put_page(skb_frag_page(frag)); ++ net_put_page(skb_frag_page(frag)); + } + + /** +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/Kconfig ./net/Kconfig +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/Kconfig 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/Kconfig 2014-08-07 09:14:18.606676153 +0200 +@@ -74,6 +74,18 @@ config INET + + Short answer: say Y. + ++config TCP_ZERO_COPY_TRANSFER_COMPLETION_NOTIFICATION ++ bool "TCP/IP zero-copy transfer completion notification" ++ depends on INET ++ default SCST_ISCSI ++ ---help--- ++ Adds support for sending a notification upon completion of a ++ zero-copy TCP/IP transfer. This can speed up certain TCP/IP ++ software. Currently this is only used by the iSCSI target driver ++ iSCSI-SCST. ++ ++ If unsure, say N. ++ + if INET + source "net/ipv4/Kconfig" + source "net/ipv6/Kconfig" +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ceph/pagevec.c ./net/ceph/pagevec.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ceph/pagevec.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/ceph/pagevec.c 2014-08-07 09:14:18.606676153 +0200 +@@ -51,7 +51,7 @@ void ceph_put_page_vector(struct page ** + for (i = 0; i < num_pages; i++) { + if (dirty) + set_page_dirty_lock(pages[i]); +- put_page(pages[i]); ++ net_put_page(pages[i]); + } + kfree(pages); + } +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/core/skbuff.c ./net/core/skbuff.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/core/skbuff.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/core/skbuff.c 2014-08-07 09:14:18.606676153 +0200 +@@ -453,7 +453,7 @@ struct sk_buff *__netdev_alloc_skb(struc + if (likely(data)) { + skb = build_skb(data, fragsz); + if (unlikely(!skb)) +- put_page(virt_to_head_page(data)); ++ net_put_page(virt_to_head_page(data)); + } + } else { + skb = __alloc_skb(length + NET_SKB_PAD, gfp_mask, +@@ -499,7 +499,7 @@ static void skb_clone_fraglist(struct sk + static void skb_free_head(struct sk_buff *skb) + { + if (skb->head_frag) +- put_page(virt_to_head_page(skb->head)); ++ net_put_page(virt_to_head_page(skb->head)); + else + kfree(skb->head); + } +@@ -824,7 +824,7 @@ int skb_copy_ubufs(struct sk_buff *skb, + if (!page) { + while (head) { + struct page *next = (struct page *)head->private; +- put_page(head); ++ net_put_page(head); + head = next; + } + return -ENOMEM; +@@ -1644,7 +1644,7 @@ EXPORT_SYMBOL(skb_copy_bits); + */ + static void sock_spd_release(struct splice_pipe_desc *spd, unsigned int i) + { +- put_page(spd->pages[i]); ++ net_put_page(spd->pages[i]); + } + + static struct page *linear_to_page(struct page *page, unsigned int *len, +@@ -1697,7 +1697,7 @@ static bool spd_fill_page(struct splice_ + spd->partial[spd->nr_pages - 1].len += *len; + return false; + } +- get_page(page); ++ net_get_page(page); + spd->pages[spd->nr_pages] = page; + spd->partial[spd->nr_pages].len = *len; + spd->partial[spd->nr_pages].offset = offset; +@@ -2794,7 +2794,7 @@ int skb_append_datato_frags(struct sock + copy); + frg_cnt++; + pfrag->offset += copy; +- get_page(pfrag->page); ++ net_get_page(pfrag->page); + + skb->truesize += copy; + atomic_add(copy, &sk->sk_wmem_alloc); +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/core/sock.c ./net/core/sock.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/core/sock.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/core/sock.c 2014-08-07 09:14:18.606676153 +0200 +@@ -1829,7 +1829,7 @@ bool sk_page_frag_refill(struct sock *sk + } + if (pfrag->offset < pfrag->size) + return true; +- put_page(pfrag->page); ++ net_put_page(pfrag->page); + } + + /* We restrict high order allocations to users that can afford to wait */ +@@ -2534,7 +2534,7 @@ void sk_common_release(struct sock *sk) + sk_refcnt_debug_release(sk); + + if (sk->sk_frag.page) { +- put_page(sk->sk_frag.page); ++ net_put_page(sk->sk_frag.page); + sk->sk_frag.page = NULL; + } + +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/Makefile ./net/ipv4/Makefile +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/Makefile 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/ipv4/Makefile 2014-08-07 09:14:18.606676153 +0200 +@@ -53,6 +53,7 @@ obj-$(CONFIG_TCP_CONG_YEAH) += tcp_yeah. + obj-$(CONFIG_TCP_CONG_ILLINOIS) += tcp_illinois.o + obj-$(CONFIG_MEMCG_KMEM) += tcp_memcontrol.o + obj-$(CONFIG_NETLABEL) += cipso_ipv4.o ++obj-$(CONFIG_TCP_ZERO_COPY_TRANSFER_COMPLETION_NOTIFICATION) += tcp_zero_copy.o + + obj-$(CONFIG_XFRM) += xfrm4_policy.o xfrm4_state.o xfrm4_input.o \ + xfrm4_output.o +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/ip_output.c ./net/ipv4/ip_output.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/ip_output.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/ipv4/ip_output.c 2014-08-07 09:14:18.606676153 +0200 +@@ -1011,7 +1011,7 @@ alloc_new_skb: + __skb_fill_page_desc(skb, i, pfrag->page, + pfrag->offset, 0); + skb_shinfo(skb)->nr_frags = ++i; +- get_page(pfrag->page); ++ net_get_page(pfrag->page); + } + copy = min_t(int, copy, pfrag->size - pfrag->offset); + if (getfrag(from, +@@ -1235,7 +1235,7 @@ ssize_t ip_append_page(struct sock *sk, + if (skb_can_coalesce(skb, i, page, offset)) { + skb_frag_size_add(&skb_shinfo(skb)->frags[i-1], len); + } else if (i < MAX_SKB_FRAGS) { +- get_page(page); ++ net_get_page(page); + skb_fill_page_desc(skb, i, page, offset, len); + } else { + err = -EMSGSIZE; +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/tcp.c ./net/ipv4/tcp.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/tcp.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/ipv4/tcp.c 2014-08-07 09:14:18.606676153 +0200 +@@ -898,7 +898,7 @@ new_segment: + if (can_coalesce) { + skb_frag_size_add(&skb_shinfo(skb)->frags[i - 1], copy); + } else { +- get_page(page); ++ net_get_page(page); + skb_fill_page_desc(skb, i, page, offset, copy); + } + skb_shinfo(skb)->tx_flags |= SKBTX_SHARED_FRAG; +@@ -1194,7 +1194,7 @@ new_segment: + } else { + skb_fill_page_desc(skb, i, pfrag->page, + pfrag->offset, copy); +- get_page(pfrag->page); ++ net_get_page(pfrag->page); + } + pfrag->offset += copy; + } +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/tcp_zero_copy.c ./net/ipv4/tcp_zero_copy.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv4/tcp_zero_copy.c 1970-01-01 01:00:00.000000000 +0100 ++++ ./net/ipv4/tcp_zero_copy.c 2014-08-07 09:14:18.606676153 +0200 +@@ -0,0 +1,50 @@ ++/* ++ * Support routines for TCP zero copy transmit ++ * ++ * Created by Vladislav Bolkhovitin ++ * ++ * This program is free software; you can redistribute it and/or ++ * modify it under the terms of the GNU General Public License ++ * version 2 as published by the Free Software Foundation. ++ */ ++ ++#include ++#include ++ ++net_get_page_callback_t net_get_page_callback __read_mostly; ++EXPORT_SYMBOL_GPL(net_get_page_callback); ++ ++net_put_page_callback_t net_put_page_callback __read_mostly; ++EXPORT_SYMBOL_GPL(net_put_page_callback); ++ ++/* ++ * Caller of this function must ensure that at the moment when it's called ++ * there are no pages in the system with net_priv field set to non-zero ++ * value. Hence, this function, as well as net_get_page() and net_put_page(), ++ * don't need any protection. ++ */ ++int net_set_get_put_page_callbacks( ++ net_get_page_callback_t get_callback, ++ net_put_page_callback_t put_callback) ++{ ++ int res = 0; ++ ++ if ((net_get_page_callback != NULL) && (get_callback != NULL) && ++ (net_get_page_callback != get_callback)) { ++ res = -EBUSY; ++ goto out; ++ } ++ ++ if ((net_put_page_callback != NULL) && (put_callback != NULL) && ++ (net_put_page_callback != put_callback)) { ++ res = -EBUSY; ++ goto out; ++ } ++ ++ net_get_page_callback = get_callback; ++ net_put_page_callback = put_callback; ++ ++out: ++ return res; ++} ++EXPORT_SYMBOL_GPL(net_set_get_put_page_callbacks); +diff -Nrup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv6/ip6_output.c ./net/ipv6/ip6_output.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/net/ipv6/ip6_output.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./net/ipv6/ip6_output.c 2014-08-07 09:14:18.606676153 +0200 +@@ -1421,7 +1421,7 @@ alloc_new_skb: + __skb_fill_page_desc(skb, i, pfrag->page, + pfrag->offset, 0); + skb_shinfo(skb)->nr_frags = ++i; +- get_page(pfrag->page); ++ net_get_page(pfrag->page); + } + copy = min_t(int, copy, pfrag->size - pfrag->offset); + if (getfrag(from, diff --git a/scst/kernel/rhel/scst_exec_req_fifo-3.10.0-123.el7.patch b/scst/kernel/rhel/scst_exec_req_fifo-3.10.0-123.el7.patch new file mode 100644 index 000000000..d60ddd1de --- /dev/null +++ b/scst/kernel/rhel/scst_exec_req_fifo-3.10.0-123.el7.patch @@ -0,0 +1,524 @@ +diff -rup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/block/blk-map.c ./block/blk-map.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/block/blk-map.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./block/blk-map.c 2014-08-07 09:09:11.751302961 +0200 +@@ -5,6 +5,8 @@ + #include + #include + #include ++#include ++#include + #include /* for struct sg_iovec */ + + #include "blk.h" +@@ -275,6 +277,337 @@ int blk_rq_unmap_user(struct bio *bio) + } + EXPORT_SYMBOL(blk_rq_unmap_user); + ++struct blk_kern_sg_work { ++ atomic_t bios_inflight; ++ struct sg_table sg_table; ++ struct scatterlist *src_sgl; ++}; ++ ++static void blk_free_kern_sg_work(struct blk_kern_sg_work *bw) ++{ ++ struct sg_table *sgt = &bw->sg_table; ++ struct scatterlist *sg; ++ int i; ++ ++ for_each_sg(sgt->sgl, sg, sgt->orig_nents, i) { ++ struct page *pg = sg_page(sg); ++ if (pg == NULL) ++ break; ++ __free_page(pg); ++ } ++ ++ sg_free_table(sgt); ++ kfree(bw); ++ return; ++} ++ ++static void blk_bio_map_kern_endio(struct bio *bio, int err) ++{ ++ struct blk_kern_sg_work *bw = bio->bi_private; ++ ++ if (bw != NULL) { ++ /* Decrement the bios in processing and, if zero, free */ ++ BUG_ON(atomic_read(&bw->bios_inflight) <= 0); ++ if (atomic_dec_and_test(&bw->bios_inflight)) { ++ if ((bio_data_dir(bio) == READ) && (err == 0)) { ++ unsigned long flags; ++ ++ local_irq_save(flags); /* to protect KMs */ ++ sg_copy(bw->src_sgl, bw->sg_table.sgl, 0, 0); ++ local_irq_restore(flags); ++ } ++ blk_free_kern_sg_work(bw); ++ } ++ } ++ ++ bio_put(bio); ++ return; ++} ++ ++static int blk_rq_copy_kern_sg(struct request *rq, struct scatterlist *sgl, ++ int nents, struct blk_kern_sg_work **pbw, ++ gfp_t gfp, gfp_t page_gfp) ++{ ++ int res = 0, i; ++ struct scatterlist *sg; ++ struct scatterlist *new_sgl; ++ int new_sgl_nents; ++ size_t len = 0, to_copy; ++ struct blk_kern_sg_work *bw; ++ ++ bw = kzalloc(sizeof(*bw), gfp); ++ if (bw == NULL) ++ goto out; ++ ++ bw->src_sgl = sgl; ++ ++ for_each_sg(sgl, sg, nents, i) ++ len += sg->length; ++ to_copy = len; ++ ++ new_sgl_nents = PFN_UP(len); ++ ++ res = sg_alloc_table(&bw->sg_table, new_sgl_nents, gfp); ++ if (res != 0) ++ goto err_free; ++ ++ new_sgl = bw->sg_table.sgl; ++ ++ for_each_sg(new_sgl, sg, new_sgl_nents, i) { ++ struct page *pg; ++ ++ pg = alloc_page(page_gfp); ++ if (pg == NULL) ++ goto err_free; ++ ++ sg_assign_page(sg, pg); ++ sg->length = min_t(size_t, PAGE_SIZE, len); ++ ++ len -= PAGE_SIZE; ++ } ++ ++ if (rq_data_dir(rq) == WRITE) { ++ /* ++ * We need to limit amount of copied data to to_copy, because ++ * sgl might have the last element in sgl not marked as last in ++ * SG chaining. ++ */ ++ sg_copy(new_sgl, sgl, 0, to_copy); ++ } ++ ++ *pbw = bw; ++ /* ++ * REQ_COPY_USER name is misleading. It should be something like ++ * REQ_HAS_TAIL_SPACE_FOR_PADDING. ++ */ ++ rq->cmd_flags |= REQ_COPY_USER; ++ ++out: ++ return res; ++ ++err_free: ++ blk_free_kern_sg_work(bw); ++ res = -ENOMEM; ++ goto out; ++} ++ ++static int __blk_rq_map_kern_sg(struct request *rq, struct scatterlist *sgl, ++ int nents, struct blk_kern_sg_work *bw, gfp_t gfp) ++{ ++ int res; ++ struct request_queue *q = rq->q; ++ int rw = rq_data_dir(rq); ++ int max_nr_vecs, i; ++ size_t tot_len; ++ bool need_new_bio; ++ struct scatterlist *sg, *prev_sg = NULL; ++ struct bio *bio = NULL, *hbio = NULL, *tbio = NULL; ++ int bios; ++ ++ if (unlikely((sgl == NULL) || (sgl->length == 0) || (nents <= 0))) { ++ WARN_ON(1); ++ res = -EINVAL; ++ goto out; ++ } ++ ++ /* ++ * Let's keep each bio allocation inside a single page to decrease ++ * probability of failure. ++ */ ++ max_nr_vecs = min_t(size_t, ++ ((PAGE_SIZE - sizeof(struct bio)) / sizeof(struct bio_vec)), ++ BIO_MAX_PAGES); ++ ++ need_new_bio = true; ++ tot_len = 0; ++ bios = 0; ++ for_each_sg(sgl, sg, nents, i) { ++ struct page *page = sg_page(sg); ++ void *page_addr = page_address(page); ++ size_t len = sg->length, l; ++ size_t offset = sg->offset; ++ ++ tot_len += len; ++ prev_sg = sg; ++ ++ /* ++ * Each segment must be aligned on DMA boundary and ++ * not on stack. The last one may have unaligned ++ * length as long as the total length is aligned to ++ * DMA padding alignment. ++ */ ++ if (i == nents - 1) ++ l = 0; ++ else ++ l = len; ++ if (((sg->offset | l) & queue_dma_alignment(q)) || ++ (page_addr && object_is_on_stack(page_addr + sg->offset))) { ++ res = -EINVAL; ++ goto out_free_bios; ++ } ++ ++ while (len > 0) { ++ size_t bytes; ++ int rc; ++ ++ if (need_new_bio) { ++ bio = bio_kmalloc(gfp, max_nr_vecs); ++ if (bio == NULL) { ++ res = -ENOMEM; ++ goto out_free_bios; ++ } ++ ++ if (rw == WRITE) ++ bio->bi_rw |= REQ_WRITE; ++ ++ bios++; ++ bio->bi_private = bw; ++ bio->bi_end_io = blk_bio_map_kern_endio; ++ ++ if (hbio == NULL) ++ hbio = tbio = bio; ++ else ++ tbio = tbio->bi_next = bio; ++ } ++ ++ bytes = min_t(size_t, len, PAGE_SIZE - offset); ++ ++ rc = bio_add_pc_page(q, bio, page, bytes, offset); ++ if (rc < bytes) { ++ if (unlikely(need_new_bio || (rc < 0))) { ++ if (rc < 0) ++ res = rc; ++ else ++ res = -EIO; ++ goto out_free_bios; ++ } else { ++ need_new_bio = true; ++ len -= rc; ++ offset += rc; ++ continue; ++ } ++ } ++ ++ need_new_bio = false; ++ offset = 0; ++ len -= bytes; ++ page = nth_page(page, 1); ++ } ++ } ++ ++ if (hbio == NULL) { ++ res = -EINVAL; ++ goto out_free_bios; ++ } ++ ++ /* Total length must be aligned on DMA padding alignment */ ++ if ((tot_len & q->dma_pad_mask) && ++ !(rq->cmd_flags & REQ_COPY_USER)) { ++ res = -EINVAL; ++ goto out_free_bios; ++ } ++ ++ if (bw != NULL) ++ atomic_set(&bw->bios_inflight, bios); ++ ++ while (hbio != NULL) { ++ bio = hbio; ++ hbio = hbio->bi_next; ++ bio->bi_next = NULL; ++ ++ blk_queue_bounce(q, &bio); ++ ++ res = blk_rq_append_bio(q, rq, bio); ++ if (unlikely(res != 0)) { ++ bio->bi_next = hbio; ++ hbio = bio; ++ /* We can have one or more bios bounced */ ++ goto out_unmap_bios; ++ } ++ } ++ ++ res = 0; ++ ++ rq->buffer = NULL; ++out: ++ return res; ++ ++out_unmap_bios: ++ blk_rq_unmap_kern_sg(rq, res); ++ ++out_free_bios: ++ while (hbio != NULL) { ++ bio = hbio; ++ hbio = hbio->bi_next; ++ bio_put(bio); ++ } ++ goto out; ++} ++ ++/** ++ * blk_rq_map_kern_sg - map kernel data to a request, for REQ_TYPE_BLOCK_PC ++ * @rq: request to fill ++ * @sgl: area to map ++ * @nents: number of elements in @sgl ++ * @gfp: memory allocation flags ++ * ++ * Description: ++ * Data will be mapped directly if possible. Otherwise a bounce ++ * buffer will be used. ++ */ ++int blk_rq_map_kern_sg(struct request *rq, struct scatterlist *sgl, ++ int nents, gfp_t gfp) ++{ ++ int res; ++ ++ res = __blk_rq_map_kern_sg(rq, sgl, nents, NULL, gfp); ++ if (unlikely(res != 0)) { ++ struct blk_kern_sg_work *bw = NULL; ++ ++ res = blk_rq_copy_kern_sg(rq, sgl, nents, &bw, ++ gfp, rq->q->bounce_gfp | gfp); ++ if (unlikely(res != 0)) ++ goto out; ++ ++ res = __blk_rq_map_kern_sg(rq, bw->sg_table.sgl, ++ bw->sg_table.nents, bw, gfp); ++ if (res != 0) { ++ blk_free_kern_sg_work(bw); ++ goto out; ++ } ++ } ++ ++ rq->buffer = NULL; ++ ++out: ++ return res; ++} ++EXPORT_SYMBOL(blk_rq_map_kern_sg); ++ ++/** ++ * blk_rq_unmap_kern_sg - unmap a request with kernel sg ++ * @rq: request to unmap ++ * @err: non-zero error code ++ * ++ * Description: ++ * Unmap a rq previously mapped by blk_rq_map_kern_sg(). Must be called ++ * only in case of an error! ++ */ ++void blk_rq_unmap_kern_sg(struct request *rq, int err) ++{ ++ struct bio *bio = rq->bio; ++ ++ while (bio) { ++ struct bio *b = bio; ++ bio = bio->bi_next; ++ b->bi_end_io(b, err); ++ } ++ rq->bio = NULL; ++ ++ return; ++} ++EXPORT_SYMBOL(blk_rq_unmap_kern_sg); ++ + /** + * blk_rq_map_kern - map kernel data to a request, for REQ_TYPE_BLOCK_PC usage + * @q: request queue where request should be inserted +diff -rup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/blkdev.h ./include/linux/blkdev.h +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/blkdev.h 2014-07-16 20:25:31.000000000 +0200 ++++ ./include/linux/blkdev.h 2014-08-07 09:09:11.751302961 +0200 +@@ -719,6 +719,8 @@ extern unsigned long blk_max_low_pfn, bl + #define BLK_DEFAULT_SG_TIMEOUT (60 * HZ) + #define BLK_MIN_SG_TIMEOUT (7 * HZ) + ++#define SCSI_EXEC_REQ_FIFO_DEFINED ++ + #ifdef CONFIG_BOUNCE + extern int init_emergency_isa_pool(void); + extern void blk_queue_bounce(struct request_queue *q, struct bio **bio); +@@ -838,6 +840,9 @@ extern int blk_rq_map_kern(struct reques + extern int blk_rq_map_user_iov(struct request_queue *, struct request *, + struct rq_map_data *, struct sg_iovec *, int, + unsigned int, gfp_t); ++extern int blk_rq_map_kern_sg(struct request *rq, struct scatterlist *sgl, ++ int nents, gfp_t gfp); ++extern void blk_rq_unmap_kern_sg(struct request *rq, int err); + extern int blk_execute_rq(struct request_queue *, struct gendisk *, + struct request *, int); + extern void blk_execute_rq_nowait(struct request_queue *, struct gendisk *, +diff -rup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/scatterlist.h ./include/linux/scatterlist.h +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/include/linux/scatterlist.h 2014-07-16 20:25:31.000000000 +0200 ++++ ./include/linux/scatterlist.h 2014-08-07 09:09:11.751302961 +0200 +@@ -8,6 +8,7 @@ + #include + #include + #include ++#include + + struct sg_table { + struct scatterlist *sgl; /* the list */ +@@ -244,6 +245,9 @@ size_t sg_copy_from_buffer(struct scatte + size_t sg_copy_to_buffer(struct scatterlist *sgl, unsigned int nents, + void *buf, size_t buflen); + ++int sg_copy(struct scatterlist *dst_sg, struct scatterlist *src_sg, ++ int nents_to_copy, size_t copy_len); ++ + /* + * Maximum number of entries that will be allocated in one piece, if + * a list larger than this is required then chaining will be utilized. +diff -rup ../../centos-7-orig/linux-3.10.0-123.6.3.el7/lib/scatterlist.c ./lib/scatterlist.c +--- ../../centos-7-orig/linux-3.10.0-123.6.3.el7/lib/scatterlist.c 2014-07-16 20:25:31.000000000 +0200 ++++ ./lib/scatterlist.c 2014-08-07 09:09:11.751302961 +0200 +@@ -628,3 +628,126 @@ size_t sg_copy_to_buffer(struct scatterl + return sg_copy_buffer(sgl, nents, buf, buflen, 1); + } + EXPORT_SYMBOL(sg_copy_to_buffer); ++ ++/* ++ * Can switch to the next dst_sg element, so, to copy to strictly only ++ * one dst_sg element, it must be either last in the chain, or ++ * copy_len == dst_sg->length. ++ */ ++static int sg_copy_elem(struct scatterlist **pdst_sg, size_t *pdst_len, ++ size_t *pdst_offs, struct scatterlist *src_sg, ++ size_t copy_len) ++{ ++ int res = 0; ++ struct scatterlist *dst_sg; ++ size_t src_len, dst_len, src_offs, dst_offs; ++ struct page *src_page, *dst_page; ++ ++ dst_sg = *pdst_sg; ++ dst_len = *pdst_len; ++ dst_offs = *pdst_offs; ++ dst_page = sg_page(dst_sg); ++ ++ src_page = sg_page(src_sg); ++ src_len = src_sg->length; ++ src_offs = src_sg->offset; ++ ++ do { ++ void *saddr, *daddr; ++ size_t n; ++ ++ saddr = kmap_atomic(src_page + (src_offs >> PAGE_SHIFT)) + ++ (src_offs & ~PAGE_MASK); ++ daddr = kmap_atomic(dst_page + (dst_offs >> PAGE_SHIFT)) + ++ (dst_offs & ~PAGE_MASK); ++ ++ if (((src_offs & ~PAGE_MASK) == 0) && ++ ((dst_offs & ~PAGE_MASK) == 0) && ++ (src_len >= PAGE_SIZE) && (dst_len >= PAGE_SIZE) && ++ (copy_len >= PAGE_SIZE)) { ++ copy_page(daddr, saddr); ++ n = PAGE_SIZE; ++ } else { ++ n = min_t(size_t, PAGE_SIZE - (dst_offs & ~PAGE_MASK), ++ PAGE_SIZE - (src_offs & ~PAGE_MASK)); ++ n = min(n, src_len); ++ n = min(n, dst_len); ++ n = min_t(size_t, n, copy_len); ++ memcpy(daddr, saddr, n); ++ } ++ dst_offs += n; ++ src_offs += n; ++ ++ kunmap_atomic(saddr); ++ kunmap_atomic(daddr); ++ ++ res += n; ++ copy_len -= n; ++ if (copy_len == 0) ++ goto out; ++ ++ src_len -= n; ++ dst_len -= n; ++ if (dst_len == 0) { ++ dst_sg = sg_next(dst_sg); ++ if (dst_sg == NULL) ++ goto out; ++ dst_page = sg_page(dst_sg); ++ dst_len = dst_sg->length; ++ dst_offs = dst_sg->offset; ++ } ++ } while (src_len > 0); ++ ++out: ++ *pdst_sg = dst_sg; ++ *pdst_len = dst_len; ++ *pdst_offs = dst_offs; ++ return res; ++} ++ ++/** ++ * sg_copy - copy one SG vector to another ++ * @dst_sg: destination SG ++ * @src_sg: source SG ++ * @nents_to_copy: maximum number of entries to copy ++ * @copy_len: maximum amount of data to copy. If 0, then copy all. ++ * ++ * Description: ++ * Data from the source SG vector will be copied to the destination SG ++ * vector. End of the vectors will be determined by sg_next() returning ++ * NULL. Returns number of bytes copied. ++ */ ++int sg_copy(struct scatterlist *dst_sg, struct scatterlist *src_sg, ++ int nents_to_copy, size_t copy_len) ++{ ++ int res = 0; ++ size_t dst_len, dst_offs; ++ ++ if (copy_len == 0) ++ copy_len = 0x7FFFFFFF; /* copy all */ ++ ++ if (nents_to_copy == 0) ++ nents_to_copy = 0x7FFFFFFF; /* copy all */ ++ ++ dst_len = dst_sg->length; ++ dst_offs = dst_sg->offset; ++ ++ do { ++ int copied = sg_copy_elem(&dst_sg, &dst_len, &dst_offs, ++ src_sg, copy_len); ++ copy_len -= copied; ++ res += copied; ++ if ((copy_len == 0) || (dst_sg == NULL)) ++ goto out; ++ ++ nents_to_copy--; ++ if (nents_to_copy == 0) ++ goto out; ++ ++ src_sg = sg_next(src_sg); ++ } while (src_sg != NULL); ++ ++out: ++ return res; ++} ++EXPORT_SYMBOL(sg_copy); +