| Message ID | 20220131155846.32411-3-idryomov@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | libceph: rxbounce support | expand |
On Mon, 2022-01-31 at 16:58 +0100, Ilya Dryomov wrote: > 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_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 > Signed-off-by: Ilya Dryomov <idryomov@gmail.com> > --- > 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(-) > > diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h > index 6a89ea410e43..edf62eaa6285 100644 > --- a/include/linux/ceph/libceph.h > +++ b/include/linux/ceph/libceph.h > @@ -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) > > diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h > index 6c6b6ea52bb8..e7f2fb2fc207 100644 > --- a/include/linux/ceph/messenger.h > +++ b/include/linux/ceph/messenger.h > @@ -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 */ > diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c > index ecc400a0b7bb..4c6441536d55 100644 > --- a/net/ceph/ceph_common.c > +++ b/net/ceph/ceph_common.c > @@ -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), 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? > 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,", > diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c > index 45eba2dcb67a..d3bb656308b4 100644 > --- a/net/ceph/messenger.c > +++ b/net/ceph/messenger.c > @@ -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); > diff --git a/net/ceph/messenger_v1.c b/net/ceph/messenger_v1.c > index 2cb5ffdf071a..6b014eca3a13 100644 > --- a/net/ceph/messenger_v1.c > +++ b/net/ceph/messenger_v1.c > @@ -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; > } > diff --git a/net/ceph/messenger_v2.c b/net/ceph/messenger_v2.c > index d34349f112b0..8a101f5d4c23 100644 > --- a/net/ceph/messenger_v2.c > +++ b/net/ceph/messenger_v2.c > @@ -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) > @@ -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); Reviewed-by: Jeff Layton <jlayton@kernel.org>
On Tue, Feb 1, 2022 at 2:24 PM Jeff Layton <jlayton@kernel.org> wrote: > > On Mon, 2022-01-31 at 16:58 +0100, Ilya Dryomov wrote: > > 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_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 > > Signed-off-by: Ilya Dryomov <idryomov@gmail.com> > > --- > > 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(-) > > > > diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h > > index 6a89ea410e43..edf62eaa6285 100644 > > --- a/include/linux/ceph/libceph.h > > +++ b/include/linux/ceph/libceph.h > > @@ -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) > > > > diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h > > index 6c6b6ea52bb8..e7f2fb2fc207 100644 > > --- a/include/linux/ceph/messenger.h > > +++ b/include/linux/ceph/messenger.h > > @@ -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 */ > > diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c > > index ecc400a0b7bb..4c6441536d55 100644 > > --- a/net/ceph/ceph_common.c > > +++ b/net/ceph/ceph_common.c > > @@ -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), > > 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. 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. > > 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... [1] https://github.com/ceph/ceph/pull/44842/files#diff-90064d9b1388dcb61bb57ff61f60443b15ef3d74f31fef7f63e43b5ae3a03f37 Thanks, Ilya
On Tue, 2022-02-01 at 16:46 +0100, Ilya Dryomov wrote: > On Tue, Feb 1, 2022 at 2:24 PM Jeff Layton <jlayton@kernel.org> wrote: > > > > On Mon, 2022-01-31 at 16:58 +0100, Ilya Dryomov wrote: > > > 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_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 > > > Signed-off-by: Ilya Dryomov <idryomov@gmail.com> > > > --- > > > 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(-) > > > > > > diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h > > > index 6a89ea410e43..edf62eaa6285 100644 > > > --- a/include/linux/ceph/libceph.h > > > +++ b/include/linux/ceph/libceph.h > > > @@ -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) > > > > > > diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h > > > index 6c6b6ea52bb8..e7f2fb2fc207 100644 > > > --- a/include/linux/ceph/messenger.h > > > +++ b/include/linux/ceph/messenger.h > > > @@ -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 */ > > > diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c > > > index ecc400a0b7bb..4c6441536d55 100644 > > > --- a/net/ceph/ceph_common.c > > > +++ b/net/ceph/ceph_common.c > > > @@ -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), > > > > 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. > > 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. > > > > 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... > 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. > [1] https://github.com/ceph/ceph/pull/44842/files#diff-90064d9b1388dcb61bb57ff61f60443b15ef3d74f31fef7f63e43b5ae3a03f37 > > Thanks, > > Ilya
diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h index 6a89ea410e43..edf62eaa6285 100644 --- a/include/linux/ceph/libceph.h +++ b/include/linux/ceph/libceph.h @@ -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) diff --git a/include/linux/ceph/messenger.h b/include/linux/ceph/messenger.h index 6c6b6ea52bb8..e7f2fb2fc207 100644 --- a/include/linux/ceph/messenger.h +++ b/include/linux/ceph/messenger.h @@ -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 */ diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c index ecc400a0b7bb..4c6441536d55 100644 --- a/net/ceph/ceph_common.c +++ b/net/ceph/ceph_common.c @@ -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,", diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c index 45eba2dcb67a..d3bb656308b4 100644 --- a/net/ceph/messenger.c +++ b/net/ceph/messenger.c @@ -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); diff --git a/net/ceph/messenger_v1.c b/net/ceph/messenger_v1.c index 2cb5ffdf071a..6b014eca3a13 100644 --- a/net/ceph/messenger_v1.c +++ b/net/ceph/messenger_v1.c @@ -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; } diff --git a/net/ceph/messenger_v2.c b/net/ceph/messenger_v2.c index d34349f112b0..8a101f5d4c23 100644 --- a/net/ceph/messenger_v2.c +++ b/net/ceph/messenger_v2.c @@ -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) @@ -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);
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_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 Signed-off-by: Ilya Dryomov <idryomov@gmail.com> --- 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(-)