libceph: rxbounce support
Ilya Dryomov
maintainercover libceph: rxbounce support
Hello,
See the second patch for rationale and explanation. Tests are added in
https://github.com/ceph/ceph/pull/44842.
Thanks,
Ilya
Ilya Dryomov (2):
libceph: make recv path in secure mode work the same as send path
libceph: optionally use bounce buffer on recv path in crc mode
include/linux/ceph/libceph.h | 1 + include/linux/ceph/messenger.h | 5 + net/ceph/ceph_common.c | 7 + net/ceph/messenger.c | 4 + net/ceph/messenger_v1.c | 54 ++++++- net/ceph/messenger_v2.c | 265 +++++++++++++++++++++++---------- 6 files changed, 255 insertions(+), 81 deletions(-)
Signature · 1 line
2.19.2
Ilya Dryomov
maintainerPATCH 1/2 libceph: make recv path in secure mode work the same as send path
The recv path of secure mode is intertwined with that of crc mode. While it's slightly more efficient that way (the ciphertext is read into the destination buffer and decrypted in place, thus avoiding two potentially heavy memory allocations for the bounce buffer and the corresponding sg array), it isn't really amenable to changes. Sacrifice that edge and align with the send path which always uses a full-sized bounce buffer (currently there is no other way -- if the kernel crypto API ever grows support for streaming (piecewise) en/decryption for GCM [1], we would be able to easily take advantage of that on both sides).
[1] https://lore.kernel.org/all/[email protected]/
--- include/linux/ceph/messenger.h | 4 + net/ceph/messenger_v2.c | 231 ++++++++++++++++++++++----------- 2 files changed, 162 insertions(+), 73 deletions(-)
include/linux/ceph/messenger.h +4 -0
@@ -383,6 +383,10 @@ struct ceph_connection_v2_info {struct ceph_gcm_nonce in_gcm_nonce;struct ceph_gcm_nonce out_gcm_nonce; +struct page **in_enc_pages; +int in_enc_page_cnt; +int in_enc_resid; +int in_enc_i;struct page **out_enc_pages;int out_enc_page_cnt;int out_enc_resid;
net/ceph/messenger_v2.c +158 -73
350 lines not on this page · raw message
Signature · 1 line
2.19.2
Re: [PATCH 1/2] libceph: make recv path in secure mode work the same as send path
Quoted 392 lines · Ilya Dryomov
The recv path of secure mode is intertwined with that of crc mode. While it's slightly more efficient that way (the ciphertext is read into the destination buffer and decrypted in place, thus avoiding two potentially heavy memory allocations for the bounce buffer and the corresponding sg array), it isn't really amenable to changes. Sacrifice that edge and align with the send path which always uses a full-sized bounce buffer (currently there is no other way -- if the kernel crypto API ever grows support for streaming (piecewise) en/decryption for GCM [1], we would be able to easily take advantage of that on both sides).
[1] https://lore.kernel.org/all/[email protected]/
Signed-off-by Ilya Dryomov <[email protected]>--- include/linux/ceph/messenger.h | 4 + net/ceph/messenger_v2.c | 231 ++++++++++++++++++++++----------- 2 files changed, 162 insertions(+), 73 deletions(-)include/linux/ceph/messenger.h +4
-0@@ -383,6 +383,10 @@ struct ceph_connection_v2_info {struct ceph_gcm_nonce in_gcm_nonce;struct ceph_gcm_nonce out_gcm_nonce; +struct page **in_enc_pages; +int in_enc_page_cnt; +int in_enc_resid; +int in_enc_i;struct page **out_enc_pages;int out_enc_page_cnt;int out_enc_resid;net/ceph/messenger_v2.c +151
-73@@ -57,8 +57,9 @@#define IN_S_HANDLE_CONTROL_REMAINDER 3#define IN_S_PREPARE_READ_DATA 4#define IN_S_PREPARE_READ_DATA_CONT 5 -#define IN_S_HANDLE_EPILOGUE 6 -#define IN_S_FINISH_SKIP 7 +#define IN_S_PREPARE_READ_ENC_PAGE 6 +#define IN_S_HANDLE_EPILOGUE 7 +#define IN_S_FINISH_SKIP 8#define OUT_S_QUEUE_DATA 1#define OUT_S_QUEUE_DATA_CONT 2@@ -1032,22 +1033,41 @@ static int decrypt_control_remainder(struct ceph_connection *con) padded_len(rem_len) + CEPH_GCM_TAG_LEN); } -static int decrypt_message (struct ceph_connection *con) +static int decrypt_tail (struct ceph_connection *con) { +struct sg_table enc_sgt = {};struct sg_table sgt = {}; +int tail_len;int ret; + tail_len =tail_onwire_len (con->in_msg, true); + ret =sg_alloc_table_from_pages (&enc_sgt, con->v2.in_enc_pages, + con->v2.in_enc_page_cnt,0 , tail_len, +GFP_NOIO ); +if (ret) + goto out; + ret =setup_message_sgs (&sgt, con->in_msg,FRONT_PAD (con->v2.in_buf),MIDDLE_PAD (con->v2.in_buf),DATA_PAD (con->v2.in_buf), con->v2.in_buf, true);if (ret) goto out; - ret =gcm_crypt (con, false, sgt.sgl, sgt.sgl, -tail_onwire_len (con->in_msg, true)); +dout ("%s con %p msg %p enc_page_cnt %d sg_cnt %d\n" , __func__, con, + con->in_msg, con->v2.in_enc_page_cnt, sgt.orig_nents); + ret =gcm_crypt (con, false, enc_sgt.sgl, sgt.sgl, tail_len); +if (ret) + goto out; + +WARN_ON (!con->v2.in_enc_page_cnt); +ceph_release_page_vector (con->v2.in_enc_pages, + con->v2.in_enc_page_cnt); + con->v2.in_enc_pages =NULL ; + con->v2.in_enc_page_cnt =0 ; out:sg_free_table (&sgt); +sg_free_table (&enc_sgt);return ret; }@@ -1737,8 +1757,7 @@ static void prepare_read_data(struct ceph_connection *con) {struct bio_vec bv; -if (!con_secure (con)) - con->in_data_crc =-1 ; + con->in_data_crc =-1 ;ceph_msg_data_cursor_init (&con->v2.in_cursor, con->in_msg,data_len (con->in_msg));@@ -1751,11 +1770,10 @@ static void prepare_read_data_cont(struct ceph_connection *con) {struct bio_vec bv; -if (!con_secure (con)) - con->in_data_crc =ceph_crc32c_page (con->in_data_crc, - con->v2.in_bvec.bv_page, - con->v2.in_bvec.bv_offset, - con->v2.in_bvec.bv_len); + con->in_data_crc =ceph_crc32c_page (con->in_data_crc, + con->v2.in_bvec.bv_page, + con->v2.in_bvec.bv_offset, + con->v2.in_bvec.bv_len);ceph_msg_data_advance (&con->v2.in_cursor, con->v2.in_bvec.bv_len);if (con->v2.in_cursor.total_resid) {@@ -1766,21 +1784,100 @@ static void prepare_read_data_cont(struct ceph_connection *con) }/* -* We've read all data. Prepare to read data padding (if any) -* and epilogue. +* We've read all data. Prepare to read epilogue. */ reset_in_kvecs (con); -if (con_secure (con)) { -if (need_padding (data_len (con->in_msg))) -add_in_kvec (con,DATA_PAD (con->v2.in_buf), -padding_len (data_len (con->in_msg))); -add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_SECURE_LEN ); +add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_PLAIN_LEN ); + con->v2.in_state =IN_S_HANDLE_EPILOGUE ; +} + +static void prepare_read_tail_plain (struct ceph_connection *con) +{ +struct ceph_msg *msg = con->in_msg; + +if (!front_len (msg) && !middle_len (msg)) { +WARN_ON (!data_len (msg)); +prepare_read_data (con); +return ; + } + +reset_in_kvecs (con); +if (front_len (msg)) { +WARN_ON (front_len (msg) > msg->front_alloc_len); +add_in_kvec (con, msg->front.iov_base,front_len (msg)); + msg->front.iov_len =front_len (msg); + }else { + msg->front.iov_len =0 ; + } +if (middle_len (msg)) { +WARN_ON (middle_len (msg) > msg->middle->alloc_len); +add_in_kvec (con, msg->middle->vec.iov_base,middle_len (msg)); + msg->middle->vec.iov_len =middle_len (msg); + }else if (msg->middle) { + msg->middle->vec.iov_len =0 ; + } + +if (data_len (msg)) { + con->v2.in_state =IN_S_PREPARE_READ_DATA ; }else {add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_PLAIN_LEN ); + con->v2.in_state =IN_S_HANDLE_EPILOGUE ; + } +} + +static void prepare_read_enc_page (struct ceph_connection *con) +{ +struct bio_vec bv; + +dout ("%s con %p i %d resid %d\n" , __func__, con, con->v2.in_enc_i, + con->v2.in_enc_resid); +WARN_ON (!con->v2.in_enc_resid); + + bv.bv_page = con->v2.in_enc_pages[con->v2.in_enc_i]; + bv.bv_offset =0 ; + bv.bv_len =min (con->v2.in_enc_resid, (int )PAGE_SIZE ); + +set_in_bvec (con, &bv); + con->v2.in_enc_i++; + con->v2.in_enc_resid -= bv.bv_len; + +if (con->v2.in_enc_resid) { + con->v2.in_state =IN_S_PREPARE_READ_ENC_PAGE ; +return ; } + +/* +* We are set to read the last piece of ciphertext (ending +* with epilogue) + auth tag. +*/ +WARN_ON (con->v2.in_enc_i != con->v2.in_enc_page_cnt); con->v2.in_state =IN_S_HANDLE_EPILOGUE ; } +static int prepare_read_tail_secure (struct ceph_connection *con) +{ +struct page **enc_pages; +int enc_page_cnt; +int tail_len; + + tail_len =tail_onwire_len (con->in_msg, true); +WARN_ON (!tail_len); + + enc_page_cnt =calc_pages_for (0 , tail_len); + enc_pages =ceph_alloc_page_vector (enc_page_cnt,GFP_NOIO ); +if (IS_ERR (enc_pages)) +return PTR_ERR (enc_pages); + +WARN_ON (con->v2.in_enc_pages || con->v2.in_enc_page_cnt); + con->v2.in_enc_pages = enc_pages; + con->v2.in_enc_page_cnt = enc_page_cnt; + con->v2.in_enc_resid = tail_len; + con->v2.in_enc_i =0 ; + +prepare_read_enc_page (con); +return 0 ; +} +static void __finish_skip (struct ceph_connection *con) { con->in_seq++;@@ -2589,46 +2686,13 @@ static int __handle_control(struct ceph_connection *con, void *p) } msg = con->in_msg;/* set in process_message_header() */ -if (!front_len (msg) && !middle_len (msg)) { -if (!data_len (msg)) -return process_message (con); - -prepare_read_data (con); -return 0 ; - } - -reset_in_kvecs (con); -if (front_len (msg)) { -WARN_ON (front_len (msg) > msg->front_alloc_len); -add_in_kvec (con, msg->front.iov_base,front_len (msg)); - msg->front.iov_len =front_len (msg); - -if (con_secure (con) &&need_padding (front_len (msg))) -add_in_kvec (con,FRONT_PAD (con->v2.in_buf), -padding_len (front_len (msg))); - }else { - msg->front.iov_len =0 ; - } -if (middle_len (msg)) { -WARN_ON (middle_len (msg) > msg->middle->alloc_len); -add_in_kvec (con, msg->middle->vec.iov_base,middle_len (msg)); - msg->middle->vec.iov_len =middle_len (msg); +if (!front_len (msg) && !middle_len (msg) && !data_len (msg)) +return process_message (con); -if (con_secure (con) &&need_padding (middle_len (msg))) -add_in_kvec (con,MIDDLE_PAD (con->v2.in_buf), -padding_len (middle_len (msg))); - }else if (msg->middle) { - msg->middle->vec.iov_len =0 ; - } +if (con_secure (con)) +return prepare_read_tail_secure (con); -if (data_len (msg)) { - con->v2.in_state =IN_S_PREPARE_READ_DATA ; - }else { -add_in_kvec (con, con->v2.in_buf, -con_secure (con) ?CEPH_EPILOGUE_SECURE_LEN : -CEPH_EPILOGUE_PLAIN_LEN ); - con->v2.in_state =IN_S_HANDLE_EPILOGUE ; - } +prepare_read_tail_plain (con);return 0 ; }@@ -2717,7 +2781,7 @@ static int handle_epilogue(struct ceph_connection *con)int ret;if (con_secure (con)) { - ret =decrypt_message (con); + ret =decrypt_tail (con);if (ret) {if (ret == -EBADMSG ) con->error_msg ="integrity error, bad epilogue auth tag" ;@@ -2792,6 +2856,10 @@ static int populate_in_iter(struct ceph_connection *con)prepare_read_data_cont (con); ret =0 ;break ; +case IN_S_PREPARE_READ_ENC_PAGE : +prepare_read_enc_page (con); + ret =0 ; +break ;case IN_S_HANDLE_EPILOGUE : ret =handle_epilogue (con);break ;@@ -3326,20 +3394,16 @@ void ceph_con_v2_revoke(struct ceph_connection *con)static void revoke_at_prepare_read_data (struct ceph_connection *con) { -int remaining;/* data + [data padding] + epilogue */ +int remaining;int resid; +WARN_ON (con_secure (con));WARN_ON (!data_len (con->in_msg));WARN_ON (!iov_iter_is_kvec (&con->v2.in_iter)); resid =iov_iter_count (&con->v2.in_iter);WARN_ON (!resid); -if (con_secure (con)) - remaining =padded_len (data_len (con->in_msg)) + -CEPH_EPILOGUE_SECURE_LEN ; -else - remaining =data_len (con->in_msg) +CEPH_EPILOGUE_PLAIN_LEN ; - + remaining =data_len (con->in_msg) +CEPH_EPILOGUE_PLAIN_LEN ;dout ("%s con %p resid %d remaining %d\n" , __func__, con, resid, remaining); con->v2.in_iter.count -= resid;@@ -3350,8 +3414,9 @@ static void revoke_at_prepare_read_data(struct ceph_connection *con)static void revoke_at_prepare_read_data_cont (struct ceph_connection *con) {int recved, resid;/* current piece of data */ -int remaining;/* [data padding] + epilogue */ +int remaining; +WARN_ON (con_secure (con));WARN_ON (!data_len (con->in_msg));WARN_ON (!iov_iter_is_bvec (&con->v2.in_iter)); resid =iov_iter_count (&con->v2.in_iter);@@ -3363,12 +3428,7 @@ static void revoke_at_prepare_read_data_cont(struct ceph_connection *con)ceph_msg_data_advance (&con->v2.in_cursor, recved);WARN_ON (resid > con->v2.in_cursor.total_resid); -if (con_secure (con)) - remaining =padding_len (data_len (con->in_msg)) + -CEPH_EPILOGUE_SECURE_LEN ; -else - remaining =CEPH_EPILOGUE_PLAIN_LEN ; - + remaining =CEPH_EPILOGUE_PLAIN_LEN ;dout ("%s con %p total_resid %zu remaining %d\n" , __func__, con, con->v2.in_cursor.total_resid, remaining); con->v2.in_iter.count -= resid;@@ -3376,11 +3436,26 @@ static void revoke_at_prepare_read_data_cont(struct ceph_connection *con) con->v2.in_state =IN_S_FINISH_SKIP ; } +static void revoke_at_prepare_read_enc_page (struct ceph_connection *con) +{ +int resid;/* current enc page (not necessarily data) */ + +WARN_ON (!con_secure (con)); +WARN_ON (!iov_iter_is_bvec (&con->v2.in_iter)); + resid =iov_iter_count (&con->v2.in_iter); +WARN_ON (!resid || resid > con->v2.in_bvec.bv_len); + +dout ("%s con %p resid %d enc_resid %d\n" , __func__, con, resid, + con->v2.in_enc_resid); + con->v2.in_iter.count -= resid; +set_in_skip (con, resid + con->v2.in_enc_resid); + con->v2.in_state =IN_S_FINISH_SKIP ; +} +static void revoke_at_handle_epilogue (struct ceph_connection *con) {int resid; -WARN_ON (!iov_iter_is_kvec (&con->v2.in_iter)); resid =iov_iter_count (&con->v2.in_iter);WARN_ON (!resid);@@ -3399,6 +3474,9 @@ void ceph_con_v2_revoke_incoming(struct ceph_connection *con)case IN_S_PREPARE_READ_DATA_CONT :revoke_at_prepare_read_data_cont (con);break ; +case IN_S_PREPARE_READ_ENC_PAGE : +revoke_at_prepare_read_enc_page (con); +break ;case IN_S_HANDLE_EPILOGUE :revoke_at_handle_epilogue (con);break ;@@ -3432,6 +3510,13 @@ void ceph_con_v2_reset_protocol(struct ceph_connection *con)clear_out_sign_kvecs (con);free_conn_bufs (con);
net/ceph/messenger_v2.c void ceph_con_v2_reset_protocol(struct ceph_connection *con) · line 3522
Quoted 10 lines
net/ceph/messenger_v2.c +7
-0+if (con->v2.in_enc_pages) { +WARN_ON (!con->v2.in_enc_page_cnt); +ceph_release_page_vector (con->v2.in_enc_pages, + con->v2.in_enc_page_cnt); + con->v2.in_enc_pages =NULL ; + con->v2.in_enc_page_cnt =0 ; + }if (con->v2.out_enc_pages) {WARN_ON (!con->v2.out_enc_page_cnt); ceph_release_page_vector(con->v2.out_enc_pages,
Ilya Dryomov
maintainerRe: [PATCH 1/2] libceph: make recv path in secure mode work the same as send path
In reply to Ilya DryomovQuoted 259 lines · Ilya Dryomov
The recv path of secure mode is intertwined with that of crc mode. While it's slightly more efficient that way (the ciphertext is read into the destination buffer and decrypted in place, thus avoiding two potentially heavy memory allocations for the bounce buffer and the corresponding sg array), it isn't really amenable to changes. Sacrifice that edge and align with the send path which always uses a full-sized bounce buffer (currently there is no other way -- if the kernel crypto API ever grows support for streaming (piecewise) en/decryption for GCM [1], we would be able to easily take advantage of that on both sides).
[1] https://lore.kernel.org/all/[email protected]/
Signed-off-by Ilya Dryomov <[email protected]>--- include/linux/ceph/messenger.h | 4 + net/ceph/messenger_v2.c | 231 ++++++++++++++++++++++----------- 2 files changed, 162 insertions(+), 73 deletions(-)include/linux/ceph/messenger.h +4
-0@@ -383,6 +383,10 @@ struct ceph_connection_v2_info {struct ceph_gcm_nonce in_gcm_nonce;struct ceph_gcm_nonce out_gcm_nonce; +struct page **in_enc_pages; +int in_enc_page_cnt; +int in_enc_resid; +int in_enc_i;struct page **out_enc_pages;int out_enc_page_cnt;int out_enc_resid;net/ceph/messenger_v2.c +117
-43@@ -57,8 +57,9 @@#define IN_S_HANDLE_CONTROL_REMAINDER 3#define IN_S_PREPARE_READ_DATA 4#define IN_S_PREPARE_READ_DATA_CONT 5 -#define IN_S_HANDLE_EPILOGUE 6 -#define IN_S_FINISH_SKIP 7 +#define IN_S_PREPARE_READ_ENC_PAGE 6 +#define IN_S_HANDLE_EPILOGUE 7 +#define IN_S_FINISH_SKIP 8#define OUT_S_QUEUE_DATA 1#define OUT_S_QUEUE_DATA_CONT 2@@ -1032,22 +1033,41 @@ static int decrypt_control_remainder(struct ceph_connection *con) padded_len(rem_len) + CEPH_GCM_TAG_LEN); } -static int decrypt_message (struct ceph_connection *con) +static int decrypt_tail (struct ceph_connection *con) { +struct sg_table enc_sgt = {};struct sg_table sgt = {}; +int tail_len;int ret; + tail_len =tail_onwire_len (con->in_msg, true); + ret =sg_alloc_table_from_pages (&enc_sgt, con->v2.in_enc_pages, + con->v2.in_enc_page_cnt,0 , tail_len, +GFP_NOIO ); +if (ret) + goto out; + ret =setup_message_sgs (&sgt, con->in_msg,FRONT_PAD (con->v2.in_buf),MIDDLE_PAD (con->v2.in_buf),DATA_PAD (con->v2.in_buf), con->v2.in_buf, true);if (ret) goto out; - ret =gcm_crypt (con, false, sgt.sgl, sgt.sgl, -tail_onwire_len (con->in_msg, true)); +dout ("%s con %p msg %p enc_page_cnt %d sg_cnt %d\n" , __func__, con, + con->in_msg, con->v2.in_enc_page_cnt, sgt.orig_nents); + ret =gcm_crypt (con, false, enc_sgt.sgl, sgt.sgl, tail_len); +if (ret) + goto out; + +WARN_ON (!con->v2.in_enc_page_cnt); +ceph_release_page_vector (con->v2.in_enc_pages, + con->v2.in_enc_page_cnt); + con->v2.in_enc_pages =NULL ; + con->v2.in_enc_page_cnt =0 ; out:sg_free_table (&sgt); +sg_free_table (&enc_sgt);return ret; }@@ -1737,8 +1757,7 @@ static void prepare_read_data(struct ceph_connection *con) {struct bio_vec bv; -if (!con_secure (con)) - con->in_data_crc =-1 ; + con->in_data_crc =-1 ;ceph_msg_data_cursor_init (&con->v2.in_cursor, con->in_msg,data_len (con->in_msg));@@ -1751,11 +1770,10 @@ static void prepare_read_data_cont(struct ceph_connection *con) {struct bio_vec bv; -if (!con_secure (con)) - con->in_data_crc =ceph_crc32c_page (con->in_data_crc, - con->v2.in_bvec.bv_page, - con->v2.in_bvec.bv_offset, - con->v2.in_bvec.bv_len); + con->in_data_crc =ceph_crc32c_page (con->in_data_crc, + con->v2.in_bvec.bv_page, + con->v2.in_bvec.bv_offset, + con->v2.in_bvec.bv_len);ceph_msg_data_advance (&con->v2.in_cursor, con->v2.in_bvec.bv_len);if (con->v2.in_cursor.total_resid) {@@ -1766,21 +1784,100 @@ static void prepare_read_data_cont(struct ceph_connection *con) }/* -* We've read all data. Prepare to read data padding (if any) -* and epilogue. +* We've read all data. Prepare to read epilogue. */ reset_in_kvecs (con); -if (con_secure (con)) { -if (need_padding (data_len (con->in_msg))) -add_in_kvec (con,DATA_PAD (con->v2.in_buf), -padding_len (data_len (con->in_msg))); -add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_SECURE_LEN ); +add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_PLAIN_LEN ); + con->v2.in_state =IN_S_HANDLE_EPILOGUE ; +} + +static void prepare_read_tail_plain (struct ceph_connection *con) +{ +struct ceph_msg *msg = con->in_msg; + +if (!front_len (msg) && !middle_len (msg)) { +WARN_ON (!data_len (msg)); +prepare_read_data (con); +return ; + } + +reset_in_kvecs (con); +if (front_len (msg)) { +WARN_ON (front_len (msg) > msg->front_alloc_len); +add_in_kvec (con, msg->front.iov_base,front_len (msg)); + msg->front.iov_len =front_len (msg); + }else { + msg->front.iov_len =0 ; + } +if (middle_len (msg)) { +WARN_ON (middle_len (msg) > msg->middle->alloc_len); +add_in_kvec (con, msg->middle->vec.iov_base,middle_len (msg)); + msg->middle->vec.iov_len =middle_len (msg); + }else if (msg->middle) { + msg->middle->vec.iov_len =0 ; + } + +if (data_len (msg)) { + con->v2.in_state =IN_S_PREPARE_READ_DATA ; }else {add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_PLAIN_LEN ); + con->v2.in_state =IN_S_HANDLE_EPILOGUE ; + } +} + +static void prepare_read_enc_page (struct ceph_connection *con) +{ +struct bio_vec bv; + +dout ("%s con %p i %d resid %d\n" , __func__, con, con->v2.in_enc_i, + con->v2.in_enc_resid); +WARN_ON (!con->v2.in_enc_resid); + + bv.bv_page = con->v2.in_enc_pages[con->v2.in_enc_i]; + bv.bv_offset =0 ; + bv.bv_len =min (con->v2.in_enc_resid, (int )PAGE_SIZE ); + +set_in_bvec (con, &bv); + con->v2.in_enc_i++; + con->v2.in_enc_resid -= bv.bv_len; + +if (con->v2.in_enc_resid) { + con->v2.in_state =IN_S_PREPARE_READ_ENC_PAGE ; +return ; } + +/* +* We are set to read the last piece of ciphertext (ending +* with epilogue) + auth tag. +*/ +WARN_ON (con->v2.in_enc_i != con->v2.in_enc_page_cnt); con->v2.in_state =IN_S_HANDLE_EPILOGUE ; } +static int prepare_read_tail_secure (struct ceph_connection *con) +{ +struct page **enc_pages; +int enc_page_cnt; +int tail_len; + + tail_len =tail_onwire_len (con->in_msg, true); +WARN_ON (!tail_len); + + enc_page_cnt =calc_pages_for (0 , tail_len); + enc_pages =ceph_alloc_page_vector (enc_page_cnt,GFP_NOIO ); +if (IS_ERR (enc_pages)) +return PTR_ERR (enc_pages); + +WARN_ON (con->v2.in_enc_pages || con->v2.in_enc_page_cnt); + con->v2.in_enc_pages = enc_pages; + con->v2.in_enc_page_cnt = enc_page_cnt; + con->v2.in_enc_resid = tail_len; + con->v2.in_enc_i =0 ; + +prepare_read_enc_page (con); +return 0 ; +} +static void __finish_skip (struct ceph_connection *con) { con->in_seq++;@@ -2589,46 +2686,13 @@ static int __handle_control(struct ceph_connection *con, void *p) } msg = con->in_msg;/* set in process_message_header() */ -if (!front_len (msg) && !middle_len (msg)) { -if (!data_len (msg)) -return process_message (con); - -prepare_read_data (con); -return 0 ; - } - -reset_in_kvecs (con); -if (front_len (msg)) { -WARN_ON (front_len (msg) > msg->front_alloc_len); -add_in_kvec (con, msg->front.iov_base,front_len (msg)); - msg->front.iov_len =front_len (msg); - -if (con_secure (con) &&need_padding (front_len (msg))) -add_in_kvec (con,FRONT_PAD (con->v2.in_buf), -padding_len (front_len (msg))); - }else { - msg->front.iov_len =0 ; - } -if (middle_len (msg)) { -WARN_ON (middle_len (msg) > msg->middle->alloc_len); -add_in_kvec (con, msg->middle->vec.iov_base,middle_len (msg)); - msg->middle->vec.iov_len =middle_len (msg); +if (!front_len (msg) && !middle_len (msg) && !data_len (msg))
net/ceph/messenger_v2.c static int __handle_control(struct ceph_connection *con, void *p) · line 2693
Quoted 10 lines
net/ceph/messenger_v2.c +3
-6+return process_message (con); -if (con_secure (con) &&need_padding (middle_len (msg))) -add_in_kvec (con,MIDDLE_PAD (con->v2.in_buf), -padding_len (middle_len (msg))); - }else if (msg->middle) { - msg->middle->vec.iov_len =0 ; - } +if (con_secure (con)) +return prepare_read_tail_secure (con);
This isn't quite right: the front and/or middle iov_len isn't set in secure mode. In most cases it's irrelevant but it does cause issues with code like:
void *p = buf;
void *end = buf + len;
...
advance p
...
if (p != end)
goto bad;I have folded in the attached incremental.
Thanks,
Ilya
net/ceph/messenger_v2.c +15 -8
@@ -1834,18 +1834,12 @@ static int prepare_read_tail_plain(struct ceph_connection *con)reset_in_kvecs (con);if (front_len (msg)) { -WARN_ON (front_len (msg) > msg->front_alloc_len);add_in_kvec (con, msg->front.iov_base,front_len (msg)); - msg->front.iov_len =front_len (msg); - }else { - msg->front.iov_len =0 ; +WARN_ON (msg->front.iov_len !=front_len (msg)); }if (middle_len (msg)) { -WARN_ON (middle_len (msg) > msg->middle->alloc_len);add_in_kvec (con, msg->middle->vec.iov_base,middle_len (msg)); - msg->middle->vec.iov_len =middle_len (msg); - }else if (msg->middle) { - msg->middle->vec.iov_len =0 ; +WARN_ON (msg->middle->vec.iov_len !=middle_len (msg)); }if (data_len (msg)) {
@@ -2718,6 +2712,19 @@ static int __handle_control(struct ceph_connection *con, void *p) } msg = con->in_msg;/* set in process_message_header() */ +if (front_len (msg)) { +WARN_ON (front_len (msg) > msg->front_alloc_len); + msg->front.iov_len =front_len (msg); + }else { + msg->front.iov_len =0 ; + } +if (middle_len (msg)) { +WARN_ON (middle_len (msg) > msg->middle->alloc_len); + msg->middle->vec.iov_len =middle_len (msg); + }else if (msg->middle) { + msg->middle->vec.iov_len =0 ; + } +if (!front_len (msg) && !middle_len (msg) && !data_len (msg))return process_message (con);
Ilya Dryomov
maintainerPATCH 2/2 libceph: optionally use bounce buffer on recv path in crc mode
In reply to Ilya DryomovBoth msgr1 and msgr2 in crc mode are zero copy in the sense that message data is read from the socket directly into the destination buffer. We assume that the destination buffer is stable (i.e. remains unchanged while it is being read to) though. Otherwise, CRC errors ensue:
libceph: read_partial_message 0000000048edf8ad data crc 1063286393 != exp. 228122706
libceph: osd1 (1)192.168.122.1:6843 bad crc/signature
libceph: bad data crc, calculated 57958023, expected 1805382778
libceph: osd2 (2)192.168.122.1:6876 integrity error, bad crc
Introduce rxbounce option to enable use of a bounce buffer when receiving message data. In particular this is needed if a mapped image is a Windows VM disk, passed to QEMU. Windows has a system-wide "dummy" page that may be mapped into the destination buffer (potentially more than once into the same buffer) by the Windows Memory Manager in an effort to generate a single large I/O [1][2]. QEMU makes a point of preserving overlap relationships when cloning I/O vectors, so krbd gets exposed to this behaviour.
[1] "What Is Really in That MDL?"
https://docs.microsoft.com/en-us/previous-versions/windows/hardware/design/dn614012(v=vs.85)
[2] https://blogs.msmvps.com/kernelmustard/2005/05/04/dummy-pages/
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1973317
--- include/linux/ceph/libceph.h | 1 + include/linux/ceph/messenger.h | 1 + net/ceph/ceph_common.c | 7 ++++ net/ceph/messenger.c | 4 +++ net/ceph/messenger_v1.c | 54 +++++++++++++++++++++++++++---- net/ceph/messenger_v2.c | 58 ++++++++++++++++++++++++++-------- 6 files changed, 105 insertions(+), 20 deletions(-)
include/linux/ceph/libceph.h +1 -0
@@ -35,6 +35,7 @@#define CEPH_OPT_TCP_NODELAY (1<<4)/* TCP_NODELAY on TCP sockets */ #define CEPH_OPT_NOMSGSIGN (1<<5)/* don't sign msgs (msgr1) */ #define CEPH_OPT_ABORT_ON_FULL (1<<6)/* abort w/ ENOSPC when full */ +#define CEPH_OPT_RXBOUNCE (1<<7)/* double-buffer read data */ #define CEPH_OPT_DEFAULT (CEPH_OPT_TCP_NODELAY)
include/linux/ceph/messenger.h +1 -0
@@ -461,6 +461,7 @@ struct ceph_connection {struct ceph_msg *out_msg;/* sending message (== tail of out_sent) */ +struct page *bounce_page;u32 in_front_crc, in_middle_crc, in_data_crc;/* calculated crc */ struct timespec64 last_keepalive_ack;/* keepalive2 ack stamp */
net/ceph/ceph_common.c +7 -0
@@ -246,6 +246,7 @@ enum { Opt_cephx_sign_messages, Opt_tcp_nodelay, Opt_abort_on_full, + Opt_rxbounce, }; enum {
@@ -295,6 +296,7 @@ static const struct fs_parameter_spec ceph_parameters[] = {fsparam_u32 ("osdkeepalive" , Opt_osdkeepalivetimeout),fsparam_enum ("read_from_replica" , Opt_read_from_replica, ceph_param_read_from_replica), +fsparam_flag ("rxbounce" , Opt_rxbounce),fsparam_enum ("ms_mode" , Opt_ms_mode, ceph_param_ms_mode), fsparam_string ("secret", Opt_secret),
@@ -584,6 +586,9 @@ int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,case Opt_abort_on_full: opt->flags |=CEPH_OPT_ABORT_ON_FULL ;break ; +case Opt_rxbounce: + opt->flags |=CEPH_OPT_RXBOUNCE ; +break ;default :BUG ();
@@ -660,6 +665,8 @@ int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,seq_puts (m,"notcp_nodelay," );if (show_all && (opt->flags &CEPH_OPT_ABORT_ON_FULL ))seq_puts (m,"abort_on_full," ); +if (opt->flags &CEPH_OPT_RXBOUNCE ) +seq_puts (m,"rxbounce," );if (opt->mount_timeout !=CEPH_MOUNT_TIMEOUT_DEFAULT ) seq_printf(m, "mount_timeout=%d,",
net/ceph/messenger.c +4 -0
@@ -515,6 +515,10 @@ static void ceph_con_reset_protocol(struct ceph_connection *con)ceph_msg_put (con->out_msg); con->out_msg =NULL ; } +if (con->bounce_page) { +__free_page (con->bounce_page); + con->bounce_page =NULL ; + }if (ceph_msgr2 (from_msgr (con->msgr)))ceph_con_v2_reset_protocol (con);
net/ceph/messenger_v1.c +48 -6
78 lines not on this page · raw message
net/ceph/messenger_v2.c +44 -14
110 lines not on this page · raw message
Signature · 1 line
2.19.2
Re: [PATCH 2/2] libceph: optionally use bounce buffer on recv path in crc mode
Quoted 73 lines · Ilya Dryomov
Both msgr1 and msgr2 in crc mode are zero copy in the sense that message data is read from the socket directly into the destination buffer. We assume that the destination buffer is stable (i.e. remains unchanged while it is being read to) though. Otherwise, CRC errors ensue:
libceph:
read_partial_message0000000048edf8addata crc 1063286393 != exp. 228122706
libceph: osd1 (1)192.168.122.1:6843 bad crc/signaturelibceph: bad data crc, calculated 57958023, expected 1805382778
libceph: osd2 (2)192.168.122.1:6876 integrity error, bad crcIntroduce rxbounce option to enable use of a bounce buffer when receiving message data. In particular this is needed if a mapped image is a Windows VM disk, passed to QEMU. Windows has a system-wide "dummy" page that may be mapped into the destination buffer (potentially more than once into the same buffer) by the Windows Memory Manager in an effort to generate a single large I/O [1][2]. QEMU makes a point of preserving overlap relationships when cloning I/O vectors, so krbd gets exposed to this behaviour.
[1] "What Is Really in That MDL?"
https://docs.microsoft.com/en-us/previous-versions/windows/hardware/design/dn614012(v=vs.85)[2] https://blogs.msmvps.com/kernelmustard/2005/05/04/dummy-pages/
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1973317
Signed-off-by Ilya Dryomov <[email protected]>--- include/linux/ceph/libceph.h | 1 + include/linux/ceph/messenger.h | 1 + net/ceph/ceph_common.c | 7 ++++ net/ceph/messenger.c | 4 +++ net/ceph/messenger_v1.c | 54 +++++++++++++++++++++++++++---- net/ceph/messenger_v2.c | 58 ++++++++++++++++++++++++++-------- 6 files changed, 105 insertions(+), 20 deletions(-)include/linux/ceph/libceph.h +1
-0@@ -35,6 +35,7 @@#define CEPH_OPT_TCP_NODELAY (1<<4)/* TCP_NODELAY on TCP sockets */ #define CEPH_OPT_NOMSGSIGN (1<<5)/* don't sign msgs (msgr1) */ #define CEPH_OPT_ABORT_ON_FULL (1<<6)/* abort w/ ENOSPC when full */ +#define CEPH_OPT_RXBOUNCE (1<<7)/* double-buffer read data */ #define CEPH_OPT_DEFAULT (CEPH_OPT_TCP_NODELAY)include/linux/ceph/messenger.h +1
-0@@ -461,6 +461,7 @@ struct ceph_connection {struct ceph_msg *out_msg;/* sending message (== tail of out_sent) */ +struct page *bounce_page;u32 in_front_crc, in_middle_crc, in_data_crc;/* calculated crc */ struct timespec64 last_keepalive_ack;/* keepalive2 ack stamp */ net/ceph/ceph_common.c +1
-0@@ -246,6 +246,7 @@ enum { Opt_cephx_sign_messages, Opt_tcp_nodelay, Opt_abort_on_full, + Opt_rxbounce, }; enum {
net/ceph/ceph_common.c static const struct fs_parameter_spec ceph_parameters[] = { · line 299
Quoted 4 lines
net/ceph/ceph_common.c +1
-0@@ -295,6 +296,7 @@ static const struct fs_parameter_spec ceph_parameters[] = {fsparam_u32 ("osdkeepalive" , Opt_osdkeepalivetimeout),fsparam_enum ("read_from_replica" , Opt_read_from_replica, ceph_param_read_from_replica), + fsparam_flag ("rxbounce", Opt_rxbounce),
Yuck.
It sure would be nice to automagically detect when this was needed somehow. The option is fine once you know you need it, but getting to that point may be painful.
Maybe we should we make the warnings about failing crc messages suggest rxbounce? We could also consider making it so that when you fail a crc check and the connection is reset, that the new connection enables rxbounce automatically?
Quoted 234 lines
net/ceph/ceph_common.c +5
-0fsparam_enum ("ms_mode" , Opt_ms_mode, ceph_param_ms_mode), fsparam_string ("secret", Opt_secret),@@ -584,6 +586,9 @@ int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,case Opt_abort_on_full: opt->flags |=CEPH_OPT_ABORT_ON_FULL ;break ; +case Opt_rxbounce: + opt->flags |=CEPH_OPT_RXBOUNCE ; +break ;default :BUG ();@@ -660,6 +665,8 @@ int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,seq_puts (m,"notcp_nodelay," );if (show_all && (opt->flags &CEPH_OPT_ABORT_ON_FULL ))seq_puts (m,"abort_on_full," ); +if (opt->flags &CEPH_OPT_RXBOUNCE ) +seq_puts (m,"rxbounce," );if (opt->mount_timeout !=CEPH_MOUNT_TIMEOUT_DEFAULT ) seq_printf(m, "mount_timeout=%d,",net/ceph/messenger.c +4
-0@@ -515,6 +515,10 @@ static void ceph_con_reset_protocol(struct ceph_connection *con)ceph_msg_put (con->out_msg); con->out_msg =NULL ; } +if (con->bounce_page) { +__free_page (con->bounce_page); + con->bounce_page =NULL ; + }if (ceph_msgr2 (from_msgr (con->msgr)))ceph_con_v2_reset_protocol (con);net/ceph/messenger_v1.c +48
-6@@ -992,8 +992,7 @@ static int read_partial_message_section(struct ceph_connection *con,static int read_partial_msg_data (struct ceph_connection *con) { -struct ceph_msg *msg = con->in_msg; -struct ceph_msg_data_cursor *cursor = &msg->cursor; +struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor;bool do_datacrc = !ceph_test_opt (from_msgr (con->msgr),NOCRC );struct page *page;size_t page_offset;@@ -1001,9 +1000,6 @@ static int read_partial_msg_data(struct ceph_connection *con)u32 crc =0 ;int ret; -if (!msg->num_data_items) -return -EIO ; -if (do_datacrc) crc = con->in_data_crc;while (cursor->total_resid) {@@ -1031,6 +1027,46 @@ static int read_partial_msg_data(struct ceph_connection *con)return 1 ;/* must return > 0 to indicate success */ } +static int read_partial_msg_data_bounce (struct ceph_connection *con) +{ +struct ceph_msg_data_cursor *cursor = &con->in_msg->cursor; +struct page *page; +size_t off, len; +u32 crc; +int ret; + +if (unlikely (!con->bounce_page)) { + con->bounce_page =alloc_page (GFP_NOIO ); +if (!con->bounce_page) { +pr_err ("failed to allocate bounce page\n" ); +return -ENOMEM ; + } + } + + crc = con->in_data_crc; +while (cursor->total_resid) { +if (!cursor->resid) { +ceph_msg_data_advance (cursor,0 ); +continue ; + } + + page =ceph_msg_data_next (cursor, &off, &len,NULL ); + ret =ceph_tcp_recvpage (con->sock, con->bounce_page,0 , len); +if (ret <=0 ) { + con->in_data_crc = crc; +return ret; + } + + crc =crc32c (crc,page_address (con->bounce_page), ret); +memcpy_to_page (page, off,page_address (con->bounce_page), ret); + +ceph_msg_data_advance (cursor, ret); + } + con->in_data_crc = crc; + +return 1 ;/* must return > 0 to indicate success */ +} +/* * read (part of) a message. */ @@ -1141,7 +1177,13 @@ static int read_partial_message(struct ceph_connection *con)/* (page) data */ if (data_len) { - ret =read_partial_msg_data (con); +if (!m->num_data_items) +return -EIO ; + +if (ceph_test_opt (from_msgr (con->msgr),RXBOUNCE )) + ret =read_partial_msg_data_bounce (con); +else + ret =read_partial_msg_data (con);if (ret <=0 )return ret; }net/ceph/messenger_v2.c +43
-12@@ -1753,7 +1753,7 @@ static int prepare_read_control_remainder(struct ceph_connection *con)return 0 ; } -static void prepare_read_data (struct ceph_connection *con) +static int prepare_read_data (struct ceph_connection *con) {struct bio_vec bv;@@ -1762,23 +1762,55 @@ static void prepare_read_data(struct ceph_connection *con) data_len(con->in_msg));get_bvec_at (&con->v2.in_cursor, &bv); -set_in_bvec (con, &bv); +if (ceph_test_opt (from_msgr (con->msgr),RXBOUNCE )) { +if (unlikely (!con->bounce_page)) { + con->bounce_page =alloc_page (GFP_NOIO ); +if (!con->bounce_page) { +pr_err ("failed to allocate bounce page\n" ); +return -ENOMEM ; + } + } + + bv.bv_page = con->bounce_page; + bv.bv_offset =0 ; +set_in_bvec (con, &bv); + }else { +set_in_bvec (con, &bv); + } con->v2.in_state =IN_S_PREPARE_READ_DATA_CONT ; +return 0 ; }static void prepare_read_data_cont (struct ceph_connection *con) {struct bio_vec bv; - con->in_data_crc =ceph_crc32c_page (con->in_data_crc, - con->v2.in_bvec.bv_page, - con->v2.in_bvec.bv_offset, - con->v2.in_bvec.bv_len); +if (ceph_test_opt (from_msgr (con->msgr),RXBOUNCE )) { + con->in_data_crc =crc32c (con->in_data_crc, +page_address (con->bounce_page), + con->v2.in_bvec.bv_len); + +get_bvec_at (&con->v2.in_cursor, &bv); +memcpy_to_page (bv.bv_page, bv.bv_offset, +page_address (con->bounce_page), + con->v2.in_bvec.bv_len); + }else { + con->in_data_crc =ceph_crc32c_page (con->in_data_crc, + con->v2.in_bvec.bv_page, + con->v2.in_bvec.bv_offset, + con->v2.in_bvec.bv_len); + }ceph_msg_data_advance (&con->v2.in_cursor, con->v2.in_bvec.bv_len);if (con->v2.in_cursor.total_resid) {get_bvec_at (&con->v2.in_cursor, &bv); -set_in_bvec (con, &bv); +if (ceph_test_opt (from_msgr (con->msgr),RXBOUNCE )) { + bv.bv_page = con->bounce_page; + bv.bv_offset =0 ; +set_in_bvec (con, &bv); + }else { +set_in_bvec (con, &bv); + }WARN_ON (con->v2.in_state !=IN_S_PREPARE_READ_DATA_CONT );return ; }@@ -1791,14 +1823,13 @@ static void prepare_read_data_cont(struct ceph_connection *con) con->v2.in_state =IN_S_HANDLE_EPILOGUE ; } -static void prepare_read_tail_plain (struct ceph_connection *con) +static int prepare_read_tail_plain (struct ceph_connection *con) {struct ceph_msg *msg = con->in_msg;if (!front_len (msg) && !middle_len (msg)) {WARN_ON (!data_len (msg)); -prepare_read_data (con); -return ; +return prepare_read_data (con); }reset_in_kvecs (con);@@ -1823,6 +1854,7 @@ static void prepare_read_tail_plain(struct ceph_connection *con)add_in_kvec (con, con->v2.in_buf,CEPH_EPILOGUE_PLAIN_LEN ); con->v2.in_state =IN_S_HANDLE_EPILOGUE ; } +return 0 ; }static void prepare_read_enc_page (struct ceph_connection *con)@@ -2692,8 +2724,7 @@ static int __handle_control(struct ceph_connection *con, void *p)if (con_secure (con))return prepare_read_tail_secure (con); -prepare_read_tail_plain (con); -return 0 ; +return prepare_read_tail_plain (con); }static int handle_preamble (struct ceph_connection *con)
net/ceph/messenger_v2.c static int populate_in_iter(struct ceph_connection *con) · line 2886
Quoted 9 lines
net/ceph/messenger_v2.c +1
-2@@ -2849,8 +2880,7 @@ static int populate_in_iter(struct ceph_connection *con) ret =handle_control_remainder (con);break ;case IN_S_PREPARE_READ_DATA : -prepare_read_data (con); - ret =0 ; + ret =prepare_read_data (con);break ;case IN_S_PREPARE_READ_DATA_CONT :prepare_read_data_cont (con);
Ilya Dryomov
maintainerRe: [PATCH 2/2] libceph: optionally use bounce buffer on recv path in crc mode
Quoted 82 lines · Jeff Layton
Quoted 73 lines · Ilya Dryomov
Both msgr1 and msgr2 in crc mode are zero copy in the sense that message data is read from the socket directly into the destination buffer. We assume that the destination buffer is stable (i.e. remains unchanged while it is being read to) though. Otherwise, CRC errors ensue:
libceph:
read_partial_message0000000048edf8addata crc 1063286393 != exp. 228122706
libceph: osd1 (1)192.168.122.1:6843 bad crc/signaturelibceph: bad data crc, calculated 57958023, expected 1805382778
libceph: osd2 (2)192.168.122.1:6876 integrity error, bad crcIntroduce rxbounce option to enable use of a bounce buffer when receiving message data. In particular this is needed if a mapped image is a Windows VM disk, passed to QEMU. Windows has a system-wide "dummy" page that may be mapped into the destination buffer (potentially more than once into the same buffer) by the Windows Memory Manager in an effort to generate a single large I/O [1][2]. QEMU makes a point of preserving overlap relationships when cloning I/O vectors, so krbd gets exposed to this behaviour.
[1] "What Is Really in That MDL?"
https://docs.microsoft.com/en-us/previous-versions/windows/hardware/design/dn614012(v=vs.85)[2] https://blogs.msmvps.com/kernelmustard/2005/05/04/dummy-pages/
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1973317
Signed-off-by Ilya Dryomov <[email protected]>--- include/linux/ceph/libceph.h | 1 + include/linux/ceph/messenger.h | 1 + net/ceph/ceph_common.c | 7 ++++ net/ceph/messenger.c | 4 +++ net/ceph/messenger_v1.c | 54 +++++++++++++++++++++++++++---- net/ceph/messenger_v2.c | 58 ++++++++++++++++++++++++++-------- 6 files changed, 105 insertions(+), 20 deletions(-)include/linux/ceph/libceph.h +1
-0@@ -35,6 +35,7 @@#define CEPH_OPT_TCP_NODELAY (1<<4)/* TCP_NODELAY on TCP sockets */ #define CEPH_OPT_NOMSGSIGN (1<<5)/* don't sign msgs (msgr1) */ #define CEPH_OPT_ABORT_ON_FULL (1<<6)/* abort w/ ENOSPC when full */ +#define CEPH_OPT_RXBOUNCE (1<<7)/* double-buffer read data */ #define CEPH_OPT_DEFAULT (CEPH_OPT_TCP_NODELAY)include/linux/ceph/messenger.h +1
-0@@ -461,6 +461,7 @@ struct ceph_connection {struct ceph_msg *out_msg;/* sending message (== tail of out_sent) */ +struct page *bounce_page;u32 in_front_crc, in_middle_crc, in_data_crc;/* calculated crc */ struct timespec64 last_keepalive_ack;/* keepalive2 ack stamp */ net/ceph/ceph_common.c +1
-0@@ -246,6 +246,7 @@ enum { Opt_cephx_sign_messages, Opt_tcp_nodelay, Opt_abort_on_full, + Opt_rxbounce, }; enum {net/ceph/ceph_common.c static const struct fs_parameter_spec ceph_parameters[] = { · line 299
Quoted 4 lines
net/ceph/ceph_common.c +1
-0@@ -295,6 +296,7 @@ static const struct fs_parameter_spec ceph_parameters[] = {fsparam_u32 ("osdkeepalive" , Opt_osdkeepalivetimeout),fsparam_enum ("read_from_replica" , Opt_read_from_replica, ceph_param_read_from_replica), + fsparam_flag ("rxbounce", Opt_rxbounce),Yuck.
Quoted 3 lines
It sure would be nice to automagically detect when this was needed somehow. The option is fine once you know you need it, but getting to that point may be painful.
I'm updating the man page in [1]. I should probably expand rxbounce paragraph with the exact error strings -- that should make it easy to find for anyone interested.
Quoted 5 lines
Maybe we should we make the warnings about failing crc messages suggest rxbounce? We could also consider making it so that when you fail a crc check and the connection is reset, that the new connection enables rxbounce automatically?
Do you envision that happening after just a single checksum mismatch? For that particular connection or globally? If we decide to do that, I think it should be global (just because currently all connection settings are global) but then some sort of threshold is going to be needed...
Thanks,
Ilya
Re: [PATCH 2/2] libceph: optionally use bounce buffer on recv path in crc mode
Quoted 83 lines · Ilya Dryomov
Quoted 82 lines · Jeff Layton
Quoted 73 lines · Ilya Dryomov
Both msgr1 and msgr2 in crc mode are zero copy in the sense that message data is read from the socket directly into the destination buffer. We assume that the destination buffer is stable (i.e. remains unchanged while it is being read to) though. Otherwise, CRC errors ensue:
libceph:
read_partial_message0000000048edf8addata crc 1063286393 != exp. 228122706
libceph: osd1 (1)192.168.122.1:6843 bad crc/signaturelibceph: bad data crc, calculated 57958023, expected 1805382778
libceph: osd2 (2)192.168.122.1:6876 integrity error, bad crcIntroduce rxbounce option to enable use of a bounce buffer when receiving message data. In particular this is needed if a mapped image is a Windows VM disk, passed to QEMU. Windows has a system-wide "dummy" page that may be mapped into the destination buffer (potentially more than once into the same buffer) by the Windows Memory Manager in an effort to generate a single large I/O [1][2]. QEMU makes a point of preserving overlap relationships when cloning I/O vectors, so krbd gets exposed to this behaviour.
[1] "What Is Really in That MDL?"
https://docs.microsoft.com/en-us/previous-versions/windows/hardware/design/dn614012(v=vs.85)[2] https://blogs.msmvps.com/kernelmustard/2005/05/04/dummy-pages/
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1973317
Signed-off-by Ilya Dryomov <[email protected]>--- include/linux/ceph/libceph.h | 1 + include/linux/ceph/messenger.h | 1 + net/ceph/ceph_common.c | 7 ++++ net/ceph/messenger.c | 4 +++ net/ceph/messenger_v1.c | 54 +++++++++++++++++++++++++++---- net/ceph/messenger_v2.c | 58 ++++++++++++++++++++++++++-------- 6 files changed, 105 insertions(+), 20 deletions(-)include/linux/ceph/libceph.h +1
-0@@ -35,6 +35,7 @@#define CEPH_OPT_TCP_NODELAY (1<<4)/* TCP_NODELAY on TCP sockets */ #define CEPH_OPT_NOMSGSIGN (1<<5)/* don't sign msgs (msgr1) */ #define CEPH_OPT_ABORT_ON_FULL (1<<6)/* abort w/ ENOSPC when full */ +#define CEPH_OPT_RXBOUNCE (1<<7)/* double-buffer read data */ #define CEPH_OPT_DEFAULT (CEPH_OPT_TCP_NODELAY)include/linux/ceph/messenger.h +1
-0@@ -461,6 +461,7 @@ struct ceph_connection {struct ceph_msg *out_msg;/* sending message (== tail of out_sent) */ +struct page *bounce_page;u32 in_front_crc, in_middle_crc, in_data_crc;/* calculated crc */ struct timespec64 last_keepalive_ack;/* keepalive2 ack stamp */ net/ceph/ceph_common.c +1
-0@@ -246,6 +246,7 @@ enum { Opt_cephx_sign_messages, Opt_tcp_nodelay, Opt_abort_on_full, + Opt_rxbounce, }; enum {net/ceph/ceph_common.c static const struct fs_parameter_spec ceph_parameters[] = { · line 299
Quoted 4 lines
net/ceph/ceph_common.c +1
-0@@ -295,6 +296,7 @@ static const struct fs_parameter_spec ceph_parameters[] = {fsparam_u32 ("osdkeepalive" , Opt_osdkeepalivetimeout),fsparam_enum ("read_from_replica" , Opt_read_from_replica, ceph_param_read_from_replica), + fsparam_flag ("rxbounce", Opt_rxbounce),Yuck.
Quoted 8 lines
Quoted 3 lines
It sure would be nice to automagically detect when this was needed somehow. The option is fine once you know you need it, but getting to that point may be painful.
I'm updating the man page in [1]. I should probably expand rxbounce paragraph with the exact error strings -- that should make it easy to find for anyone interested.
Fair enough. That seems like the reasonable place to start.
Quoted 6 lines
Quoted 5 lines
Maybe we should we make the warnings about failing crc messages suggest rxbounce? We could also consider making it so that when you fail a crc check and the connection is reset, that the new connection enables rxbounce automatically?
Quoted 6 lines
Do you envision that happening after just a single checksum mismatch? For that particular connection or globally? If we decide to do that, I think it should be global (just because currently all connection settings are global) but then some sort of threshold is going to be needed...
I don't know. I didn't really think through all of the potential issues. Probably we should just wait to see how big a problem this is before we do anything automatic.
Quoted 5 lines
Thanks,
Ilya
Signature · 1 line
Jeff Layton <[email protected]>