qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: Dima Stepanov <dimastep@yandex-team.ru>, qemu-devel@nongnu.org
Cc: wrfsh@yandex-team.ru
Subject: Re: [Qemu-devel] [PATCH v1] qemu-pr-helper: garbage response structure can be used to write data
Date: Mon, 2 Jul 2018 14:21:41 +0200	[thread overview]
Message-ID: <5149995f-2bb2-cbe8-1a2c-e16ed0ee3cd6@redhat.com> (raw)
In-Reply-To: <20180702085207.GA6032@dimastep-nix>

On 02/07/2018 10:52, Dima Stepanov wrote:
> Ping.
> 
> On Fri, Jun 15, 2018 at 12:11:44PM +0300, Dima Stepanov wrote:
>> The prh_co_entry() routine handles requests. The first part is to read a
>> request by calling the prh_read_request() routine, if:
>>   1. scsi_cdb_xfer(req->cdb) call returns 0, and
>>   2. req->cdb[0] == PERSISTENT_RESERVE_IN, then
>> The resp->result field will be uninitialized. As a result the resp.sz
>> field will be also uninitialized in the prh_co_entry() function.
>> The second part is to send the response by calling the
>> prh_write_response() routine:
>>   1. For the PERSISTENT_RESERVE_IN command, and
>>   2. resp->result == GOOD (previous successful reply or just luck), then
>> There is a probability that the following assert will not be trigered:
>>   assert(resp->sz <= req->sz && resp->sz <= sizeof(client->data));
>> As a result some uninitialized response will be sent.
>>
>> The fix is to initialize the response structure to CHECK_CONDITION and 0
>> values before calling the prh_read_request() routine.

The actual bug is that the "if (sz > 0)" should apply only to 
PERSISTENT_RESERVE_OUT, and in fact it can be done in do_pr_out.  
PERSISTENT_RESERVE_IN with sz == 0 is weird but okay.

This simplifies the code a bit too, because we can handle closing the 
file descriptor in prh_co_entry.

Does something like this work for you?

diff --git a/scsi/qemu-pr-helper.c b/scsi/qemu-pr-helper.c
index 0218d65bbf..c89a446a45 100644
--- a/scsi/qemu-pr-helper.c
+++ b/scsi/qemu-pr-helper.c
@@ -455,6 +455,14 @@ static int multipath_pr_out(int fd, const uint8_t *cdb, uint8_t *sense,
     char transportids[PR_HELPER_DATA_SIZE];
     int r;
 
+    if (sz < PR_OUT_FIXED_PARAM_SIZE) {
+        /* Illegal request, Parameter list length error.  This isn't fatal;
+         * we have read the data, send an error without closing the socket.
+         */
+        scsi_build_sense(sense, SENSE_CODE(INVALID_PARAM_LEN));
+        return CHECK_CONDITION;
+    }
+
     switch (rq_servact) {
     case MPATH_PROUT_REG_SA:
     case MPATH_PROUT_RES_SA:
@@ -574,6 +582,12 @@ static int do_pr_out(int fd, const uint8_t *cdb, uint8_t *sense,
                      const uint8_t *param, int sz)
 {
     int resp_sz;
+
+    if ((fcntl(fd, F_GETFL) & O_ACCMODE) == O_RDONLY) {
+        scsi_build_sense(sense, SENSE_CODE(INVALID_OPCODE));
+        return CHECK_CONDITION;
+    }
+
 #ifdef CONFIG_MPATH
     if (is_mpath(fd)) {
         return multipath_pr_out(fd, cdb, sense, param, sz);
@@ -690,21 +704,6 @@ static int coroutine_fn prh_read_request(PRHelperClient *client,
                                  errp) < 0) {
             goto out_close;
         }
-        if ((fcntl(client->fd, F_GETFL) & O_ACCMODE) == O_RDONLY) {
-            scsi_build_sense(resp->sense, SENSE_CODE(INVALID_OPCODE));
-            sz = 0;
-        } else if (sz < PR_OUT_FIXED_PARAM_SIZE) {
-            /* Illegal request, Parameter list length error.  This isn't fatal;
-             * we have read the data, send an error without closing the socket.
-             */
-            scsi_build_sense(resp->sense, SENSE_CODE(INVALID_PARAM_LEN));
-            sz = 0;
-        }
-        if (sz == 0) {
-            resp->result = CHECK_CONDITION;
-            close(client->fd);
-            client->fd = -1;
-        }
     }
 
     req->fd = client->fd;
@@ -785,25 +784,23 @@ static void coroutine_fn prh_co_entry(void *opaque)
             break;
         }
 
-        if (sz > 0) {
-            num_active_sockets++;
-            if (req.cdb[0] == PERSISTENT_RESERVE_OUT) {
-                r = do_pr_out(req.fd, req.cdb, resp.sense,
-                              client->data, sz);
-                resp.sz = 0;
-            } else {
-                resp.sz = sizeof(client->data);
-                r = do_pr_in(req.fd, req.cdb, resp.sense,
-                             client->data, &resp.sz);
-                resp.sz = MIN(resp.sz, sz);
-            }
-            num_active_sockets--;
-            close(req.fd);
-            if (r == -1) {
-                break;
-            }
-            resp.result = r;
+        num_active_sockets++;
+        if (req.cdb[0] == PERSISTENT_RESERVE_OUT) {
+            r = do_pr_out(req.fd, req.cdb, resp.sense,
+                          client->data, sz);
+            resp.sz = 0;
+        } else {
+            resp.sz = sizeof(client->data);
+            r = do_pr_in(req.fd, req.cdb, resp.sense,
+                         client->data, &resp.sz);
+            resp.sz = MIN(resp.sz, sz);
+        }
+        num_active_sockets--;
+        close(req.fd);
+        if (r == -1) {
+            break;
         }
+        resp.result = r;
 
         if (prh_write_response(client, &req, &resp, &local_err) < 0) {
             break;

  reply	other threads:[~2018-07-02 12:21 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-06-15  9:11 [Qemu-devel] [PATCH v1] qemu-pr-helper: garbage response structure can be used to write data Dima Stepanov
2018-07-02  8:52 ` Dima Stepanov
2018-07-02 12:21   ` Paolo Bonzini [this message]
2018-07-03  9:27     ` Dima Stepanov
2018-07-03  9:35       ` Paolo Bonzini
2018-07-02 10:38 ` Stefan Hajnoczi

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=5149995f-2bb2-cbe8-1a2c-e16ed0ee3cd6@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=dimastep@yandex-team.ru \
    --cc=qemu-devel@nongnu.org \
    --cc=wrfsh@yandex-team.ru \
    /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;
as well as URLs for NNTP newsgroup(s).