From 9da517bae11dc0a82aedc6b93c2f4032df96f689 Mon Sep 17 00:00:00 2001 From: Vladislav Bolkhovitin Date: Wed, 18 Jan 2012 23:48:02 +0000 Subject: [PATCH] Update for kernel 3.2. Netlink part in iSCSI-SCST done by Bart Van Assche git-svn-id: http://svn.code.sf.net/p/scst/svn/trunk@4077 d57e44dd-8a1f-0410-8b47-8ef2f437770f --- iscsi-scst/README | 7 +- iscsi-scst/kernel/config.c | 1 + iscsi-scst/kernel/event.c | 9 +- iscsi-scst/kernel/iscsi.h | 1 + .../patches/put_page_callback-3.2.patch | 248 ++++++++ iscsi-scst/kernel/session.c | 4 + iscsi-scst/kernel/target.c | 1 + iscsi-scst/usr/event.c | 28 +- scst/kernel/scst_exec_req_fifo-3.2.patch | 536 ++++++++++++++++++ scst/src/scst_debug.c | 9 + scst/src/scst_priv.h | 3 + 11 files changed, 830 insertions(+), 17 deletions(-) create mode 100644 iscsi-scst/kernel/patches/put_page_callback-3.2.patch create mode 100644 scst/kernel/scst_exec_req_fifo-3.2.patch diff --git a/iscsi-scst/README b/iscsi-scst/README index 9965c1f9e..0a3cea28d 100644 --- a/iscsi-scst/README +++ b/iscsi-scst/README @@ -904,10 +904,15 @@ net/core/utils.c, ignore other failures. functions. Replace them by "net_put_page" and "net_get_page" correspondingly. -That's all. Then please send your new +For vanilla kernels that should be all. Then please send your new put_page_callback-.patch to the SCST mailing list scst-devel@lists.sourceforge.net. +But some out of tree drivers (some versions of DRBD and Intel e1000 are +reported to do so) use own pages referencing on the TX path, so they +should be modified the same way as above as well. (No need to modify RX +path, because put_page_callback logic used by SCST only on the TX path.) + Credits ------- diff --git a/iscsi-scst/kernel/config.c b/iscsi-scst/kernel/config.c index f9f12dbc4..db70ac6e7 100644 --- a/iscsi-scst/kernel/config.c +++ b/iscsi-scst/kernel/config.c @@ -14,6 +14,7 @@ * GNU General Public License for more details. */ +#include #include "iscsi.h" /* Protected by target_mgmt_mutex */ diff --git a/iscsi-scst/kernel/event.c b/iscsi-scst/kernel/event.c index afb796d81..7cec5c8fa 100644 --- a/iscsi-scst/kernel/event.c +++ b/iscsi-scst/kernel/event.c @@ -17,6 +17,7 @@ * */ +#include #include #ifdef INSIDE_KERNEL_TREE #include @@ -32,7 +33,8 @@ static int event_recv_msg(struct sk_buff *skb, struct nlmsghdr *nlh) { u32 pid; - pid = NETLINK_CREDS(skb)->pid; + pid = NETLINK_CB(skb).pid; + WARN_ON(pid == 0); iscsid_pid = pid; @@ -101,15 +103,14 @@ static int __event_send(const void *buf, int buf_len) len = NLMSG_SPACE(buf_len); - skb = alloc_skb(NLMSG_SPACE(len), GFP_KERNEL); + skb = alloc_skb(len, GFP_KERNEL); if (skb == NULL) { PRINT_ERROR("alloc_skb() failed (len %d)", len); res = -ENOMEM; goto out; } - nlh = __nlmsg_put(skb, iscsid_pid, seq++, NLMSG_DONE, - len - sizeof(*nlh), 0); + nlh = __nlmsg_put(skb, iscsid_pid, seq++, NLMSG_DONE, buf_len, 0); memcpy(NLMSG_DATA(nlh), buf, buf_len); res = netlink_unicast(nl, skb, iscsid_pid, 0); diff --git a/iscsi-scst/kernel/iscsi.h b/iscsi-scst/kernel/iscsi.h index e6cf3f4a5..9d1bc556c 100644 --- a/iscsi-scst/kernel/iscsi.h +++ b/iscsi-scst/kernel/iscsi.h @@ -21,6 +21,7 @@ #include #include #include +#include #include #ifdef INSIDE_KERNEL_TREE diff --git a/iscsi-scst/kernel/patches/put_page_callback-3.2.patch b/iscsi-scst/kernel/patches/put_page_callback-3.2.patch new file mode 100644 index 000000000..6700dbc58 --- /dev/null +++ b/iscsi-scst/kernel/patches/put_page_callback-3.2.patch @@ -0,0 +1,248 @@ +=== modified file 'include/linux/mm_types.h' +--- old/include/linux/mm_types.h 2012-01-10 22:58:17 +0000 ++++ new/include/linux/mm_types.h 2012-01-10 23:17:47 +0000 +@@ -149,6 +149,17 @@ struct page { + */ + void *shadow; + #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 + } + /* + * If another subsystem starts using the double word pairing for atomic + +=== modified file 'include/linux/net.h' +--- old/include/linux/net.h 2012-01-10 22:58:17 +0000 ++++ new/include/linux/net.h 2012-01-10 23:17:47 +0000 +@@ -61,6 +61,7 @@ typedef enum { + #include /* For O_CLOEXEC and O_NONBLOCK */ + #include + #include ++#include + + struct poll_table_struct; + struct pipe_inode_info; +@@ -289,5 +290,44 @@ extern int kernel_sock_shutdown(struct s + MODULE_ALIAS("net-pf-" __stringify(pf) "-proto-" __stringify(proto) \ + "-type-" __stringify(type)) + ++#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 */ ++ + #endif /* __KERNEL__ */ + #endif /* _LINUX_NET_H */ + +=== modified file 'include/linux/skbuff.h' +--- old/include/linux/skbuff.h 2012-01-10 22:58:17 +0000 ++++ new/include/linux/skbuff.h 2012-01-10 23:17:47 +0000 +@@ -1712,7 +1712,7 @@ 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)); + } + + /** +@@ -1735,7 +1735,7 @@ 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)); + } + + /** + +=== modified file 'net/Kconfig' +--- old/net/Kconfig 2012-01-10 22:58:17 +0000 ++++ new/net/Kconfig 2012-01-10 23:17:47 +0000 +@@ -72,6 +72,18 @@ 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" + +=== modified file 'net/core/skbuff.c' +--- old/net/core/skbuff.c 2012-01-10 22:58:17 +0000 ++++ new/net/core/skbuff.c 2012-01-10 23:17:47 +0000 +@@ -77,13 +77,13 @@ static struct kmem_cache *skbuff_fclone_ + static void sock_pipe_buf_release(struct pipe_inode_info *pipe, + struct pipe_buffer *buf) + { +- put_page(buf->page); ++ net_put_page(buf->page); + } + + static void sock_pipe_buf_get(struct pipe_inode_info *pipe, + struct pipe_buffer *buf) + { +- get_page(buf->page); ++ net_get_page(buf->page); + } + + static int sock_pipe_buf_steal(struct pipe_inode_info *pipe, +@@ -654,7 +654,7 @@ 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; +@@ -1493,7 +1493,7 @@ 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 inline struct page *linear_to_page(struct page *page, unsigned int *len, +@@ -1517,7 +1517,7 @@ new_page: + off = sk->sk_sndmsg_off; + mlen = PAGE_SIZE - off; + if (mlen < 64 && mlen < *len) { +- put_page(p); ++ net_put_page(p); + goto new_page; + } + +@@ -1527,7 +1527,7 @@ new_page: + memcpy(page_address(p) + off, page_address(page) + *offset, *len); + sk->sk_sndmsg_off += *len; + *offset = off; +- get_page(p); ++ net_get_page(p); + + return p; + } +@@ -1549,7 +1549,7 @@ static inline int spd_fill_page(struct s + if (!page) + return 1; + } else +- get_page(page); ++ net_get_page(page); + + spd->pages[spd->nr_pages] = page; + spd->partial[spd->nr_pages].len = *len; + +=== modified file 'net/ipv4/Makefile' +--- old/net/ipv4/Makefile 2012-01-10 22:58:17 +0000 ++++ new/net/ipv4/Makefile 2012-01-10 23:17:47 +0000 +@@ -48,6 +48,7 @@ obj-$(CONFIG_TCP_CONG_LP) += tcp_lp.o + obj-$(CONFIG_TCP_CONG_YEAH) += tcp_yeah.o + obj-$(CONFIG_TCP_CONG_ILLINOIS) += tcp_illinois.o + obj-$(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 + +=== modified file 'net/ipv4/ip_output.c' +--- old/net/ipv4/ip_output.c 2012-01-10 22:58:17 +0000 ++++ new/net/ipv4/ip_output.c 2012-01-10 23:17:47 +0000 +@@ -1232,7 +1232,7 @@ 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; + +=== modified file 'net/ipv4/tcp.c' +--- old/net/ipv4/tcp.c 2012-01-10 22:58:17 +0000 ++++ new/net/ipv4/tcp.c 2012-01-10 23:17:47 +0000 +@@ -815,7 +815,7 @@ 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); + } + +@@ -1022,7 +1022,7 @@ new_segment: + goto new_segment; + } else if (page) { + if (off == PAGE_SIZE) { +- put_page(page); ++ net_put_page(page); + TCP_PAGE(sk) = page = NULL; + off = 0; + } +@@ -1062,9 +1062,9 @@ new_segment: + } else { + skb_fill_page_desc(skb, i, page, off, copy); + if (TCP_PAGE(sk)) { +- get_page(page); ++ net_get_page(page); + } else if (off + copy < PAGE_SIZE) { +- get_page(page); ++ net_get_page(page); + TCP_PAGE(sk) = page; + } + } + diff --git a/iscsi-scst/kernel/session.c b/iscsi-scst/kernel/session.c index b05eb4363..a9580f756 100644 --- a/iscsi-scst/kernel/session.c +++ b/iscsi-scst/kernel/session.c @@ -17,6 +17,10 @@ #include "iscsi.h" +#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 2, 0) +#include +#endif + /* target_mutex supposed to be locked */ struct iscsi_session *session_lookup(struct iscsi_target *target, u64 sid) { diff --git a/iscsi-scst/kernel/target.c b/iscsi-scst/kernel/target.c index 624451491..ffe11cfc1 100644 --- a/iscsi-scst/kernel/target.c +++ b/iscsi-scst/kernel/target.c @@ -16,6 +16,7 @@ */ #include +#include #include "iscsi.h" #include "digest.h" diff --git a/iscsi-scst/usr/event.c b/iscsi-scst/usr/event.c index 56d2389d1..89339233d 100644 --- a/iscsi-scst/usr/event.c +++ b/iscsi-scst/usr/event.c @@ -46,9 +46,9 @@ static int nl_write(int fd, void *data, int len) struct nlmsghdr nlh = {0}; iov[0].iov_base = &nlh; - iov[0].iov_len = sizeof(nlh); + iov[0].iov_len = NLMSG_HDRLEN; iov[1].iov_base = data; - iov[1].iov_len = NLMSG_SPACE(len) - sizeof(nlh); + iov[1].iov_len = NLMSG_SPACE(len) - NLMSG_HDRLEN; nlh.nlmsg_len = NLMSG_SPACE(len); nlh.nlmsg_pid = getpid(); @@ -72,21 +72,25 @@ static int nl_read(int fd, void *data, int len, bool wait) int res; iov[0].iov_base = &nlh; - iov[0].iov_len = sizeof(nlh); + iov[0].iov_len = NLMSG_HDRLEN; iov[1].iov_base = data; - iov[1].iov_len = len; + iov[1].iov_len = NLMSG_ALIGN(len); memset(&msg, 0, sizeof(msg)); - msg.msg_name= (void *)&src_addr; + msg.msg_name = (void *)&src_addr; msg.msg_namelen = sizeof(src_addr); msg.msg_iov = iov; msg.msg_iovlen = 2; res = recvmsg(fd, &msg, wait ? 0 : MSG_DONTWAIT); if (res > 0) { - res -= sizeof(nlh); + res -= NLMSG_HDRLEN; if (res < 0) res = -EPIPE; + else if (res < iov[1].iov_len) + log_error("read netlink fd (%d) error: received %d" + " bytes but expected %zd bytes (%d)", fd, res, + iov[1].iov_len, len); } return res; @@ -177,7 +181,7 @@ static int handle_e_add_target(int fd, const struct iscsi_kern_event *event) break; } - offs += rc; + offs += min((unsigned)rc, (unsigned)event->param1_size); offs += sprintf(&buf[offs], "; "); if (event->param2_size > 0) { @@ -192,7 +196,7 @@ static int handle_e_add_target(int fd, const struct iscsi_kern_event *event) } break; } - offs += rc; + offs += min((unsigned)rc, (unsigned)event->param2_size); } buf[offs] = '\0'; @@ -414,7 +418,7 @@ static int handle_e_mgmt_cmd(int fd, const struct iscsi_kern_event *event) break; } - buf[rc] = '\0'; + buf[min((unsigned)rc, (unsigned)event->param1_size)] = '\0'; log_debug(1, "Going to parse %s", buf); @@ -503,7 +507,7 @@ static int handle_e_get_attr_value(int fd, const struct iscsi_kern_event *event) break; } - buf[rc] = '\0'; + buf[min((unsigned)rc, (unsigned)event->param1_size)] = '\0'; log_debug(1, "Going to parse name %s", buf); @@ -774,7 +778,7 @@ static int handle_e_set_attr_value(int fd, const struct iscsi_kern_event *event) break; } - offs = rc; + offs = min((unsigned)rc, (unsigned)event->param1_size); offs += sprintf(&buf[offs], " "); while (1) { @@ -789,7 +793,7 @@ static int handle_e_set_attr_value(int fd, const struct iscsi_kern_event *event) break; } - offs += rc; + offs += min((unsigned)rc, (unsigned)event->param2_size); buf[offs] = '\0'; log_debug(1, "Going to parse %s", buf); diff --git a/scst/kernel/scst_exec_req_fifo-3.2.patch b/scst/kernel/scst_exec_req_fifo-3.2.patch new file mode 100644 index 000000000..2b8257ce3 --- /dev/null +++ b/scst/kernel/scst_exec_req_fifo-3.2.patch @@ -0,0 +1,536 @@ +=== modified file 'block/blk-map.c' +--- old/block/blk-map.c 2012-01-10 22:58:17 +0000 ++++ new/block/blk-map.c 2012-01-10 23:01:21 +0000 +@@ -5,6 +5,8 @@ + #include + #include + #include ++#include ++#include + #include /* for struct sg_iovec */ + + #include "blk.h" +@@ -275,6 +277,339 @@ 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, ++ KM_BIO_DST_IRQ, KM_BIO_SRC_IRQ); ++ 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, ++ KM_USER0, KM_USER1); ++ } ++ ++ *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 + +=== modified file 'include/linux/blkdev.h' +--- old/include/linux/blkdev.h 2012-01-10 22:58:17 +0000 ++++ new/include/linux/blkdev.h 2012-01-10 23:01:21 +0000 +@@ -599,6 +599,8 @@ 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); +@@ -716,6 +718,9 @@ 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 *, + +=== modified file 'include/linux/scatterlist.h' +--- old/include/linux/scatterlist.h 2012-01-10 22:58:17 +0000 ++++ new/include/linux/scatterlist.h 2012-01-10 23:01:21 +0000 +@@ -3,6 +3,7 @@ + + #include + #include ++#include + #include + #include + #include +@@ -218,6 +219,10 @@ 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, ++ enum km_type d_km_type, enum km_type s_km_type); ++ + /* + * Maximum number of entries that will be allocated in one piece, if + * a list larger than this is required then chaining will be utilized. + +=== modified file 'lib/scatterlist.c' +--- old/lib/scatterlist.c 2012-01-10 22:58:17 +0000 ++++ new/lib/scatterlist.c 2012-01-10 23:01:21 +0000 +@@ -517,3 +517,132 @@ 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, ++ enum km_type d_km_type, enum km_type s_km_type) ++{ ++ 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), s_km_type) + ++ (src_offs & ~PAGE_MASK); ++ daddr = kmap_atomic(dst_page + ++ (dst_offs >> PAGE_SHIFT), d_km_type) + ++ (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, s_km_type); ++ kunmap_atomic(daddr, d_km_type); ++ ++ 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. ++ * @d_km_type: kmap_atomic type for the destination SG ++ * @s_km_type: kmap_atomic type for the source SG ++ * ++ * 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, ++ enum km_type d_km_type, enum km_type s_km_type) ++{ ++ 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, d_km_type, s_km_type); ++ 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); + diff --git a/scst/src/scst_debug.c b/scst/src/scst_debug.c index e551b8517..5dc95668f 100644 --- a/scst/src/scst_debug.c +++ b/scst/src/scst_debug.c @@ -20,6 +20,15 @@ * GNU General Public License for more details. */ +#include + +#ifndef INSIDE_KERNEL_TREE +#include +#endif +#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 2, 0) +#include +#endif + #ifdef INSIDE_KERNEL_TREE #include #include diff --git a/scst/src/scst_priv.h b/scst/src/scst_priv.h index d3d314619..5b9717a50 100644 --- a/scst/src/scst_priv.h +++ b/scst/src/scst_priv.h @@ -22,6 +22,9 @@ #include #include +#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 2, 0) +#include +#endif #include #include #include