From: Jeff Layton <jlayton-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
To: Pavel Shilovsky <piastryyy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Cc: linux-cifs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: Re: [PATCH] CIFS: Simplify cifs_demultiplex_thread
Date: Thu, 9 Jun 2011 07:54:11 -0400 [thread overview]
Message-ID: <20110609075411.040bf5d1@corrin.poochiereds.net> (raw)
In-Reply-To: <1307609933-5632-1-git-send-email-piastryyy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
On Thu, 9 Jun 2011 12:58:52 +0400
Pavel Shilovsky <piastryyy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> Move several code blocks to separate functions:
> 1) buffer allocations,
> 2) rfc1002 header check,
> 3) read whole smb message from socket,
> 4) find mid owner.
>
> Signed-off-by: Pavel Shilovsky <piastry-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> ---
> fs/cifs/cifs_debug.c | 1 -
> fs/cifs/connect.c | 440 ++++++++++++++++++++++++++++----------------------
> 2 files changed, 247 insertions(+), 194 deletions(-)
>
Nice. This is long overdue.
That said, speaking from the experience of having to chase down subtle
bugs in this code before, I think this would be best done as a set of
patches so that we can potentially bisect in case this breaks
something. Maybe break out each function in a separate patch?
The main reason that I've not done this before is that the code doesn't
lend itself well to separate functions. You've "solved" that here with a
lot of pointer passing. That breaks up the code into separate functions
but doesn't really clean up the code. So let's consider this a good
starting point, but more work is really needed here.
> diff --git a/fs/cifs/cifs_debug.c b/fs/cifs/cifs_debug.c
> index 2fe3cf1..c4fc68a 100644
> --- a/fs/cifs/cifs_debug.c
> +++ b/fs/cifs/cifs_debug.c
> @@ -66,7 +66,6 @@ void cifs_dump_detail(struct smb_hdr *smb)
> cERROR(1, "smb buf %p len %d", smb, smbCalcSize(smb));
> }
>
> -
> void cifs_dump_mids(struct TCP_Server_Info *server)
> {
> struct list_head *tmp;
> diff --git a/fs/cifs/connect.c b/fs/cifs/connect.c
> index fb31c2c..23addcb 100644
> --- a/fs/cifs/connect.c
> +++ b/fs/cifs/connect.c
> @@ -317,24 +317,234 @@ requeue_echo:
> queue_delayed_work(system_nrt_wq, &server->echo, SMB_ECHO_INTERVAL);
> }
>
> +static struct mid_q_entry *
> +find_cifs_owner(struct TCP_Server_Info *server, struct smb_hdr *buf,
> + int *length, bool isLargeBuf, bool *isMultiRsp, char **pbigbuf)
This function name seems a little odd. Would find_cifs_mid be better?
Usually when I see a lot of pointer argument passing like this in a
function, that's a red flag to me of a poor design. In this case,
that's not really your fault as this code was already a mess. So, I'm
ok with this in principle as it might make it easier to clean up this
code later.
While we're considering overhauling this code -- perhaps, long-term we
should move this to be a state machine?
> +{
> + struct list_head *tmp, *tmp2;
> + struct mid_q_entry *mid_entry = NULL;
> +
> + spin_lock(&GlobalMid_Lock);
> + list_for_each_safe(tmp, tmp2, &server->pending_mid_q) {
> + mid_entry = list_entry(tmp, struct mid_q_entry, qhead);
> +
> + if (mid_entry->mid != buf->Mid ||
> + mid_entry->midState != MID_REQUEST_SUBMITTED ||
> + mid_entry->command != buf->Command) {
> + mid_entry = NULL;
> + continue;
> + }
> +
> + if (*length == 0 && check2ndT2(buf, server->maxBuf) > 0) {
> + /* We have a multipart transact2 resp */
> + *isMultiRsp = true;
> + if (mid_entry->resp_buf) {
> + /* merge response - fix up 1st*/
> + *length = coalesce_t2(buf, mid_entry->resp_buf);
> + if (*length > 0) {
> + *length = 0;
> + mid_entry->multiRsp = true;
> + break;
> + } else {
> + /* all parts received or
> + * packet is malformed
> + */
> + mid_entry->multiEnd = true;
> + goto multi_t2_fnd;
> + }
> + } else {
> + if (!isLargeBuf) {
> + /*
> + * FIXME: switch to already
> + * allocated largebuf?
> + */
> + cERROR(1, "1st trans2 resp "
> + "needs bigbuf");
> + } else {
> + /* Have first buffer */
> + mid_entry->resp_buf = buf;
> + mid_entry->largeBuf = true;
> + *pbigbuf = NULL;
> + }
> + }
> + break;
> + }
> + mid_entry->resp_buf = buf;
> + mid_entry->largeBuf = isLargeBuf;
> +multi_t2_fnd:
> + if (*length == 0)
> + mid_entry->midState = MID_RESPONSE_RECEIVED;
> + else
> + mid_entry->midState = MID_RESPONSE_MALFORMED;
> +#ifdef CONFIG_CIFS_STATS2
> + mid_entry->when_received = jiffies;
> +#endif
> + list_del_init(&mid_entry->qhead);
> + break;
> + }
> + spin_unlock(&GlobalMid_Lock);
> +
> + return mid_entry;
> +}
> +
> +static int
> +check_rfc1002_header(struct TCP_Server_Info *server, char *buf, int length,
> + unsigned int pdu_length, struct socket **psocket)
I don't think you really need psocket here. That should just be
server->ssocket. For that matter, pdu_length should be readable from
buf, and "length" is always going to be 4 anyway, right?
> +{
> + char temp = *buf;
> +
> + /*
> + * The first byte big endian of the length field,
> + * is actually not part of the length but the type
> + * with the most common, zero, as regular data.
> + */
> + if (temp == (char) RFC1002_SESSION_KEEP_ALIVE) {
> + return 1;
> + } else if (temp == (char)RFC1002_POSITIVE_SESSION_RESPONSE) {
> + cFYI(1, "Good RFC 1002 session rsp");
> + return 1;
> + } else if (temp == (char)RFC1002_NEGATIVE_SESSION_RESPONSE) {
> + /* we get this from Windows 98 instead of
> + an error on SMB negprot response */
> + cFYI(1, "Negative RFC1002 Session Response Error 0x%x)",
> + pdu_length);
> + /* give server a second to clean up */
> + msleep(1000);
> + /* always try 445 first on reconnect since we get NACK
> + * on some if we ever connected to port 139 (the NACK
> + * is since we do not begin with RFC1001 session
> + * initialize frame)
> + */
> + cifs_set_port((struct sockaddr *)
> + &server->dstaddr, CIFS_PORT);
> + cifs_reconnect(server);
> + *psocket = server->ssocket;
> + wake_up(&server->response_q);
> + return 1;
> + } else if (temp != (char) 0) {
> + cERROR(1, "Unknown RFC 1002 frame");
> + cifs_dump_mem(" Received Data: ", buf, length);
> + cifs_reconnect(server);
> + *psocket = server->ssocket;
> + return 1;
> + }
> +
> + /* else we have an SMB response */
> + if ((pdu_length > CIFSMaxBufSize + MAX_CIFS_HDR_SIZE - 4) ||
> + (pdu_length < sizeof(struct smb_hdr) - 1 - 4)) {
> + cERROR(1, "Invalid size SMB length %d pdu_length %d",
> + length, pdu_length+4);
> + cifs_reconnect(server);
> + *psocket = server->ssocket;
> + wake_up(&server->response_q);
> + return 1;
> + }
> +
> + return 0;
> +}
> +
> +static int
> +allocate_bufs(char **pbigbuf, char **psmallbuf, unsigned int size,
> + bool isLargeBuf)
^^^
nit: This can probably be a bool return.
> +{
> + char *bigbuf = *pbigbuf, *smallbuf = *psmallbuf;
> +
> + if (bigbuf == NULL) {
> + bigbuf = (char *)cifs_buf_get();
> + if (!bigbuf) {
> + cERROR(1, "No memory for large SMB response");
> + msleep(3000);
> + /* retry will check if exiting */
> + return 1;
> + }
> + } else if (isLargeBuf) {
> + /* we are reusing a dirty large buf, clear its start */
> + memset(bigbuf, 0, size);
> + }
> +
> + if (smallbuf == NULL) {
> + smallbuf = (char *)cifs_small_buf_get();
> + if (!smallbuf) {
> + cERROR(1, "No memory for SMB response");
> + msleep(1000);
> + /* retry will check if exiting */
> + return 1;
> + }
> + /* beginning of smb buffer is cleared in our buf_get */
> + } else /* if existing small buf clear beginning */
> + memset(smallbuf, 0, size);
> +
> + *pbigbuf = bigbuf;
> + *psmallbuf = smallbuf;
> +
> + return 0;
> +}
> +
^^^
The buffer handling could really stand to be redesigned, but splitting
it out into a separate function is a decent start.
> +static int
> +read_from_socket(struct TCP_Server_Info *server, unsigned int pdu_length,
> + struct socket **psocket, struct msghdr *smb_msg,
> + struct kvec *iov, unsigned int *ptotal_read)
> +{
> + int length, rc = 0;
> + unsigned int total_read;
> +
> + for (total_read = 0; total_read < pdu_length;
> + total_read += length) {
> + length = kernel_recvmsg(*psocket, smb_msg, iov, 1,
> + pdu_length - total_read, 0);
> + if (server->tcpStatus == CifsExiting) {
> + /* then will exit */
> + rc = 2;
> + break;
> + } else if (server->tcpStatus == CifsNeedReconnect) {
> + cifs_reconnect(server);
> + *psocket = server->ssocket;
> + /* Reconnect wakes up rspns q */
> + /* Now we will reread sock */
> + rc = 1;
> + break;
> + } else if (length == -ERESTARTSYS ||
> + length == -EAGAIN ||
> + length == -EINTR) {
> + /*
> + * Minimum sleep to prevent looping,
> + * allowing socket to clear and app
> + * threads to set tcpStatus
> + * CifsNeedReconnect if server hung.
> + */
> + usleep_range(1000, 2000);
> + length = 0;
> + continue;
> + } else if (length <= 0) {
> + cERROR(1, "Received no data, expecting %d",
> + pdu_length - total_read);
> + cifs_reconnect(server);
> + *psocket = server->ssocket;
> + rc = 1;
> + break;
> + }
> + }
> +
> + *ptotal_read = total_read;
> + return 0;
> +}
> +
> static int
> cifs_demultiplex_thread(struct TCP_Server_Info *server)
> {
> int length;
> unsigned int pdu_length, total_read;
> + char *buf = NULL, *smallbuf = NULL, *bigbuf = NULL;
> struct smb_hdr *smb_buffer = NULL;
> - struct smb_hdr *bigbuf = NULL;
> - struct smb_hdr *smallbuf = NULL;
> struct msghdr smb_msg;
> struct kvec iov;
> struct socket *csocket = server->ssocket;
> struct list_head *tmp, *tmp2;
> struct task_struct *task_to_wake = NULL;
> struct mid_q_entry *mid_entry;
> - char temp;
> bool isLargeBuf = false;
> bool isMultiRsp;
> - int reconnect;
> + int rc;
>
> current->flags |= PF_MEMALLOC;
> cFYI(1, "Demultiplex PID: %d", task_pid_nr(current));
> @@ -348,35 +558,17 @@ cifs_demultiplex_thread(struct TCP_Server_Info *server)
> while (server->tcpStatus != CifsExiting) {
> if (try_to_freeze())
> continue;
> - if (bigbuf == NULL) {
> - bigbuf = cifs_buf_get();
> - if (!bigbuf) {
> - cERROR(1, "No memory for large SMB response");
> - msleep(3000);
> - /* retry will check if exiting */
> - continue;
> - }
> - } else if (isLargeBuf) {
> - /* we are reusing a dirty large buf, clear its start */
> - memset(bigbuf, 0, sizeof(struct smb_hdr));
> - }
>
> - if (smallbuf == NULL) {
> - smallbuf = cifs_small_buf_get();
> - if (!smallbuf) {
> - cERROR(1, "No memory for SMB response");
> - msleep(1000);
> - /* retry will check if exiting */
> - continue;
> - }
> - /* beginning of smb buffer is cleared in our buf_get */
> - } else /* if existing small buf clear beginning */
> - memset(smallbuf, 0, sizeof(struct smb_hdr));
> + rc = allocate_bufs(&bigbuf, &smallbuf, sizeof(struct smb_hdr),
> + isLargeBuf);
> + if (rc)
> + continue;
>
> isLargeBuf = false;
> isMultiRsp = false;
> - smb_buffer = smallbuf;
> - iov.iov_base = smb_buffer;
> + smb_buffer = (struct smb_hdr *)smallbuf;
> + buf = smallbuf;
> + iov.iov_base = smallbuf;
> iov.iov_len = 4;
> smb_msg.msg_control = NULL;
> smb_msg.msg_controllen = 0;
> @@ -414,8 +606,7 @@ incomplete_rcv:
> allowing socket to clear and app threads to set
> tcpStatus CifsNeedReconnect if server hung */
> if (pdu_length < 4) {
> - iov.iov_base = (4 - pdu_length) +
> - (char *)smb_buffer;
> + iov.iov_base = (4 - pdu_length) + buf;
> iov.iov_len = pdu_length;
> smb_msg.msg_control = NULL;
> smb_msg.msg_controllen = 0;
Should the above code be using read_from_socket() as
well? It would be nice if we had a single function that
we always called to do reads off the socket and simply
check the return value from that. That may be reasonable
for a later patch.
> @@ -440,114 +631,42 @@ incomplete_rcv:
> /* The right amount was read from socket - 4 bytes */
> /* so we can now interpret the length field */
>
> - /* the first byte big endian of the length field,
> - is actually not part of the length but the type
> - with the most common, zero, as regular data */
> - temp = *((char *) smb_buffer);
> -
> - /* Note that FC 1001 length is big endian on the wire,
> - but we convert it here so it is always manipulated
> - as host byte order */
> + /*
> + * Note that FC 1001 length is big endian on the wire,
^^^
nit: might as well add the missing "R" here
> + * but we convert it here so it is always manipulated
> + * as host byte order
> + */
> pdu_length = be32_to_cpu(smb_buffer->smb_buf_length);
> -
> - cFYI(1, "rfc1002 length 0x%x", pdu_length+4);
> -
> - if (temp == (char) RFC1002_SESSION_KEEP_ALIVE) {
> - continue;
> - } else if (temp == (char)RFC1002_POSITIVE_SESSION_RESPONSE) {
> - cFYI(1, "Good RFC 1002 session rsp");
> - continue;
> - } else if (temp == (char)RFC1002_NEGATIVE_SESSION_RESPONSE) {
> - /* we get this from Windows 98 instead of
> - an error on SMB negprot response */
> - cFYI(1, "Negative RFC1002 Session Response Error 0x%x)",
> - pdu_length);
> - /* give server a second to clean up */
> - msleep(1000);
> - /* always try 445 first on reconnect since we get NACK
> - * on some if we ever connected to port 139 (the NACK
> - * is since we do not begin with RFC1001 session
> - * initialize frame)
> - */
> - cifs_set_port((struct sockaddr *)
> - &server->dstaddr, CIFS_PORT);
> - cifs_reconnect(server);
> - csocket = server->ssocket;
> - wake_up(&server->response_q);
> - continue;
> - } else if (temp != (char) 0) {
> - cERROR(1, "Unknown RFC 1002 frame");
> - cifs_dump_mem(" Received Data: ", (char *)smb_buffer,
> - length);
> - cifs_reconnect(server);
> - csocket = server->ssocket;
> - continue;
> - }
> -
> - /* else we have an SMB response */
> - if ((pdu_length > CIFSMaxBufSize + MAX_CIFS_HDR_SIZE - 4) ||
> - (pdu_length < sizeof(struct smb_hdr) - 1 - 4)) {
> - cERROR(1, "Invalid size SMB length %d pdu_length %d",
> - length, pdu_length+4);
> - cifs_reconnect(server);
> - csocket = server->ssocket;
> - wake_up(&server->response_q);
> + rc = check_rfc1002_header(server, buf, length, pdu_length,
> + &csocket);
> + if (rc)
> continue;
> - }
>
> - /* else length ok */
> - reconnect = 0;
> + cFYI(1, "rfc1002 length 0x%x", pdu_length+4);
>
> if (pdu_length > MAX_CIFS_SMALL_BUFFER_SIZE - 4) {
> isLargeBuf = true;
> memcpy(bigbuf, smallbuf, 4);
> - smb_buffer = bigbuf;
> + smb_buffer = (struct smb_hdr *)bigbuf;
> + buf = bigbuf;
> }
> - length = 0;
> - iov.iov_base = 4 + (char *)smb_buffer;
> +
> + iov.iov_base = 4 + buf;
> iov.iov_len = pdu_length;
> - for (total_read = 0; total_read < pdu_length;
> - total_read += length) {
> - length = kernel_recvmsg(csocket, &smb_msg, &iov, 1,
> - pdu_length - total_read, 0);
> - if (server->tcpStatus == CifsExiting) {
> - /* then will exit */
> - reconnect = 2;
> - break;
> - } else if (server->tcpStatus == CifsNeedReconnect) {
> - cifs_reconnect(server);
> - csocket = server->ssocket;
> - /* Reconnect wakes up rspns q */
> - /* Now we will reread sock */
> - reconnect = 1;
> - break;
> - } else if (length == -ERESTARTSYS ||
> - length == -EAGAIN ||
> - length == -EINTR) {
> - msleep(1); /* minimum sleep to prevent looping,
> - allowing socket to clear and app
> - threads to set tcpStatus
> - CifsNeedReconnect if server hung*/
> - length = 0;
> - continue;
> - } else if (length <= 0) {
> - cERROR(1, "Received no data, expecting %d",
> - pdu_length - total_read);
> - cifs_reconnect(server);
> - csocket = server->ssocket;
> - reconnect = 1;
> - break;
> - }
> - }
> - if (reconnect == 2)
> + rc = read_from_socket(server, pdu_length, &csocket, &smb_msg,
> + &iov, &total_read);
> + if (rc == 2)
> break;
> - else if (reconnect == 1)
> + else if (rc == 1)
> continue;
>
> total_read += 4; /* account for rfc1002 hdr */
>
> dump_smb(smb_buffer, total_read);
>
> + server->lstrp = jiffies;
> + length = 0;
> +
> /*
> * We know that we received enough to get to the MID as we
> * checked the pdu_length earlier. Now check to see
> @@ -559,75 +678,11 @@ incomplete_rcv:
> */
> length = checkSMB(smb_buffer, smb_buffer->Mid, total_read);
> if (length != 0)
> - cifs_dump_mem("Bad SMB: ", smb_buffer,
> - min_t(unsigned int, total_read, 48));
> -
> - mid_entry = NULL;
> - server->lstrp = jiffies;
> -
> - spin_lock(&GlobalMid_Lock);
> - list_for_each_safe(tmp, tmp2, &server->pending_mid_q) {
> - mid_entry = list_entry(tmp, struct mid_q_entry, qhead);
> -
> - if (mid_entry->mid != smb_buffer->Mid ||
> - mid_entry->midState != MID_REQUEST_SUBMITTED ||
> - mid_entry->command != smb_buffer->Command) {
> - mid_entry = NULL;
> - continue;
> - }
> -
> - if (length == 0 &&
> - check2ndT2(smb_buffer, server->maxBuf) > 0) {
> - /* We have a multipart transact2 resp */
> - isMultiRsp = true;
> - if (mid_entry->resp_buf) {
> - /* merge response - fix up 1st*/
> - length = coalesce_t2(smb_buffer,
> - mid_entry->resp_buf);
> - if (length > 0) {
> - length = 0;
> - mid_entry->multiRsp = true;
> - break;
> - } else {
> - /* all parts received or
> - * packet is malformed
> - */
> - mid_entry->multiEnd = true;
> - goto multi_t2_fnd;
> - }
> - } else {
> - if (!isLargeBuf) {
> - /*
> - * FIXME: switch to already
> - * allocated largebuf?
> - */
> - cERROR(1, "1st trans2 resp "
> - "needs bigbuf");
> - } else {
> - /* Have first buffer */
> - mid_entry->resp_buf =
> - smb_buffer;
> - mid_entry->largeBuf = true;
> - bigbuf = NULL;
> - }
> - }
> - break;
> - }
> - mid_entry->resp_buf = smb_buffer;
> - mid_entry->largeBuf = isLargeBuf;
> -multi_t2_fnd:
> - if (length == 0)
> - mid_entry->midState = MID_RESPONSE_RECEIVED;
> - else
> - mid_entry->midState = MID_RESPONSE_MALFORMED;
> -#ifdef CONFIG_CIFS_STATS2
> - mid_entry->when_received = jiffies;
> -#endif
> - list_del_init(&mid_entry->qhead);
> - break;
> - }
> - spin_unlock(&GlobalMid_Lock);
> + cifs_dump_mem("Bad SMB: ", buf,
> + min_t(unsigned int, total_read, 48));
>
> + mid_entry = find_cifs_owner(server, smb_buffer, &length,
> + isLargeBuf, &isMultiRsp, &bigbuf);
> if (mid_entry != NULL) {
> mid_entry->callback(mid_entry);
> /* Was previous buf put in mpx struct for multi-rsp? */
> @@ -645,13 +700,12 @@ multi_t2_fnd:
> !isMultiRsp) {
> cERROR(1, "No task to wake, unknown frame received! "
> "NumMids %d", atomic_read(&midCount));
> - cifs_dump_mem("Received Data is: ", (char *)smb_buffer,
> + cifs_dump_mem("Received Data is: ", buf,
> sizeof(struct smb_hdr));
> #ifdef CONFIG_CIFS_DEBUG2
> cifs_dump_detail(smb_buffer);
> cifs_dump_mids(server);
> #endif /* CIFS_DEBUG2 */
> -
> }
> } /* end while !EXITING */
>
--
Jeff Layton <jlayton-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
next prev parent reply other threads:[~2011-06-09 11:54 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-06-09 8:58 [PATCH] CIFS: Simplify cifs_demultiplex_thread Pavel Shilovsky
[not found] ` <1307609933-5632-1-git-send-email-piastryyy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2011-06-09 8:58 ` [PATCH] CIFS: Fix sparse error Pavel Shilovsky
[not found] ` <1307609933-5632-2-git-send-email-piastryyy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
2011-06-09 11:11 ` Jeff Layton
2011-06-09 14:13 ` Steve French
2011-06-09 11:54 ` Jeff Layton [this message]
[not found] ` <20110609075411.040bf5d1-4QP7MXygkU+dMjc06nkz3ljfA9RmPOcC@public.gmane.org>
2011-06-09 12:10 ` [PATCH] CIFS: Simplify cifs_demultiplex_thread Pavel Shilovsky
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20110609075411.040bf5d1@corrin.poochiereds.net \
--to=jlayton-h+wxahxf7alqt0dzr+alfa@public.gmane.org \
--cc=linux-cifs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=piastryyy-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox