All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Daniel P. Berrange" <berrange@redhat.com>
To: Amarnath Valluri <amarnath.valluri@intel.com>
Cc: qemu-devel@nongnu.org, patrick.ohly@intel.com,
	stefanb@linux.vnet.ibm.com, marcandre.lureau@gmail.com
Subject: Re: [Qemu-devel] [PATCH v2 9/9] tpm: Added support for TPM emulator
Date: Fri, 7 Apr 2017 15:41:00 +0100	[thread overview]
Message-ID: <20170407144100.GL26896@redhat.com> (raw)
In-Reply-To: <1491575431-32170-10-git-send-email-amarnath.valluri@intel.com>

On Fri, Apr 07, 2017 at 05:30:31PM +0300, Amarnath Valluri wrote:
> This change introduces a new TPM backend driver that can communicate with
> swtpm(software TPM emulator) using unix domain socket interface.
> 
> Swtpm uses two unix sockets, one for plain TPM commands and responses, and one
> for out-of-band control messages.
> 
> The swtpm and associated tools can be found here:
>     https://github.com/stefanberger/swtpm
> 
> Usage:
>     # setup TPM state directory
>     mkdir /tmp/mytpm
>     chown -R tss:root /tmp/mytpm
>     /usr/bin/swtpm_setup --tpm-state /tmp/mytpm --createek
> 
>     # Ask qemu to use TPM emulator with given tpm state directory
>     qemu-system-x86_64 \
>         [...] \
>         -tpmdev emulator,id=tpm0,tpmstatedir=/tmp/mytpm,logfile=/tmp/swtpm.log \
>         -device tpm-tis,tpmdev=tpm0 \
>         [...]
> 
> Signed-off-by: Amarnath Valluri <amarnath.valluri@intel.com>
> ---
>  configure             |  15 +-
>  hmp.c                 |  21 ++
>  hw/tpm/Makefile.objs  |   1 +
>  hw/tpm/tpm_emulator.c | 927 ++++++++++++++++++++++++++++++++++++++++++++++++++
>  hw/tpm/tpm_ioctl.h    | 243 +++++++++++++
>  qapi-schema.json      |  36 +-
>  qemu-options.hx       |  53 ++-
>  tpm.c                 |   2 +-
>  8 files changed, 1289 insertions(+), 9 deletions(-)
>  create mode 100644 hw/tpm/tpm_emulator.c
>  create mode 100644 hw/tpm/tpm_ioctl.h

> +static int tpm_emulator_spawn_emulator(TPMEmulator *tpm_pt)
> +{
> +    int fds[2] = { -1, -1 };
> +    int ctrl_fds[2] = { -1, -1 };
> +    pid_t cpid;
> +
> +    if (!tpm_pt->ops->has_data_path) {
> +        if (socketpair(AF_UNIX, SOCK_SEQPACKET, 0, fds) < 0) {
> +            return -1;
> +        }
> +    }
> + 
> +    if (!tpm_pt->ops->has_ctrl_path) {
> +        if (socketpair(AF_UNIX, SOCK_SEQPACKET, 0, ctrl_fds) < 0) {
> +            if (!tpm_pt->ops->has_data_path) {
> +                closesocket(fds[0]);
> +                closesocket(fds[1]);
> +            }
> +            return -1;
> +        }
> +    }
> +
> +    cpid = qemu_fork(NULL);
> +    if (cpid < 0) {
> +        error_report("tpm-emulator: Fork failure: %s", strerror(errno));
> +        if (!tpm_pt->ops->has_data_path) {
> +            closesocket(fds[0]);
> +            closesocket(fds[1]);
> +        }
> +        if (!tpm_pt->ops->has_ctrl_path) {
> +            closesocket(ctrl_fds[0]);
> +            closesocket(ctrl_fds[1]);
> +        }
> +        return -1;
> +    }
> +
> +    if (cpid == 0) { /* CHILD */
> +        enum {
> +            PARAM_PATH,
> +            PARAM_IFACE,
> +            PARAM_SERVER,  PARAM_SERVER_ARGS,
> +            PARAM_CTRL,    PARAM_CTRL_ARGS,
> +            PARAM_STATE,   PARAM_STATE_ARGS,
> +            PARAM_PIDFILE, PARAM_PIDFILE_ARGS,
> +            PARAM_LOG,     PARAM_LOG_ARGS,
> +            PARAM_MAX
> +        };
> +
> +        int i;
> +        int data_fd = -1, ctrl_fd = -1;
> +        char *argv[PARAM_MAX+1];
> +
> +        /* close all unused inherited sockets */
> +        if (fds[0] >= 0)
> +            closesocket(fds[0]);
> +        if (ctrl_fds[0] >= 0)
> +            closesocket(ctrl_fds[0]);

The 'if' checks are pointless - its already guaranteed by the
fact you check socketpair() status.

> +        i = STDERR_FILENO + 1;
> +        if (fds[1] >= 0) {
> +            data_fd = dup2(fds[1], i++);
> +            if (data_fd < 0) {
> +                error_report("tpm-emulator: dup2() failure - %s",
> +                             strerror(errno));
> +                goto exit_child;
> +            }
> +        }
> +        if (ctrl_fds[1] >= 0) {
> +            ctrl_fd = dup2(ctrl_fds[1], i++);
> +            if (ctrl_fd < 0) {
> +                error_report("tpm-emulator: dup2() failure - %s",
> +                             strerror(errno));
> +                goto exit_child;
> +            }
> +        }
> +        for ( ; i < _SC_OPEN_MAX; i++) {

Errr, _SC_OPEN_MAX is not the maximum number of FDs - it is parameter to
use with sysconf() to query the number of files - you must call sysconf().

> +            closesocket(i);

close, not closesocket - you can't assume these are all sockets.


> +        DPRINT("\n")
> +        if (execv(tpm_pt->ops->path, (char * const *)argv) < 0) {
> +            error_report("execv() failure : %s", strerror(errno));
> +        }
> +
> +exit_child:
> +        g_strfreev(argv);
> +        if (data_fd >= 0)
> +            closesocket(data_fd);
> +        if (ctrl_fd >= 0)
> +            closesocket(ctrl_fd);
> +
> +        exit(0);

You need _exit(), not exit() as you don't want to run atexit() handlers
here. You also want '1' not '0' since this is a failure scenario.

> +    } else { /* self */
> +        struct stat st;
> +        DPRINTF("child pid: %d", cpid);
> +        int rc;
> +        int timeout = 3; /* wait for max 3 seconds */
> +
> +        /* close unsed sockets */
> +        if (fds[1] >= 0)
> +            closesocket(fds[1]);
> +        if (ctrl_fds[1] >= 0)
> +            closesocket(ctrl_fds[1]);
> +
> +        while((rc = stat(TPM_EMULATOR_PIDFILE, &st)) < 0 && timeout--) {
> +            sleep(1);
> +        }

A fixed 3 second timeout will inevitably cause failures on systems with
high load.

Presumably you're trying to handle the scenario where the child process
exits without creating the pid file ?

In which case you can use 'kill(cpid, 0)' and if errno == ESRCH
then the child has exited.

> +
> +        if (timeout == -1) {
> +            error_report("tpm-emulator: failed to find pid file: %s",
> +                         strerror(errno));
> +            goto err_kill_child;
> +        }
> +
> +        tpm_pt->data_ioc = _iochannel_new(tpm_pt->ops->data_path, fds[0], NULL);
> +        if (!tpm_pt->data_ioc) {
> +            error_report("tpm-emulator: Unable to connect socket : %s",
> +                          tpm_pt->ops->data_path);
> +            goto err_kill_child;
> +        }
> +
> +        tpm_pt->ctrl_ioc = _iochannel_new(tpm_pt->ops->ctrl_path, ctrl_fds[0], NULL);
> +        if (!tpm_pt->ctrl_ioc) {
> +            error_report("tpm-emulator: Unable to connect socket : %s",
> +                          tpm_pt->ops->ctrl_path);
> +            goto err_kill_child;
> +        }
> +
> +        tpm_pt->child_running = true;
> +
> +        qemu_add_child_watch(cpid);
> +
> +        qio_channel_add_watch(tpm_pt->data_ioc, G_IO_HUP|G_IO_ERR,
> +                              tpm_emulator_fd_handler, tpm_pt, NULL);
> +    }
> +
> +    return 0;
> +
> +err_kill_child:
> +    kill(cpid, SIGTERM);
> +    closesocket(fds[0]);
> +    closesocket(ctrl_fds[0]);

You can't assume the child has gone after a SIGTERM. You need to
check, and be prepared to SIGKILL after time reasonable time,
if needed.

> +    tpm_pt->child_running = false;
> +
> +    return -1;
> +}

> +
> +static const QemuOptDesc tpm_emulator_cmdline_opts[] = {
> +    TPM_STANDARD_CMDLINE_OPTS,
> +    {
> +        .name = "tpmstatedir",
> +        .type = QEMU_OPT_STRING,
> +        .help = "TPM state directroy",
> +    },
> +    {
> +        .name = "spawn",
> +        .type = QEMU_OPT_BOOL,
> +        .help = "Wether to spwan given emlatory binary",
> +    },
> +    {
> +        .name = "path",
> +        .type = QEMU_OPT_STRING,
> +        .help = "Path to TPM emulator binary",
> +    },
> +    {
> +        .name = "data-path",
> +        .type = QEMU_OPT_STRING,
> +        .help = "Socket path to use for data exhange",
> +    },
> +    {
> +        .name = "ctrl-path",
> +        .type = QEMU_OPT_STRING,
> +        .help = "Socket path to use for out-of-band control messages",
> +    },

I'm still not convinced by the need for 2 separate UNIX sockets, unless
there's a performance reason, but that looks unlikely.


Regards,
Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://entangle-photo.org       -o-    http://search.cpan.org/~danberr/ :|

  reply	other threads:[~2017-04-07 14:41 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-04-07 14:30 [Qemu-devel] [PATCH v2 0/9] Provide support for the software TPM Amarnath Valluri
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 1/9] tpm-backend: Remove unneeded member variable from backend class Amarnath Valluri
2017-04-25 18:19   ` Stefan Berger
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 2/9] tpm-backend: Move thread handling inside TPMBackend Amarnath Valluri
2017-04-25 18:21   ` Stefan Berger
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 3/9] tpm-backend: Initialize and free data members in it's own methods Amarnath Valluri
2017-04-25 18:27   ` Stefan Berger
2017-05-02  7:09     ` Amarnath Valluri
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 4/9] tpm-backend: Made few interface methods optional Amarnath Valluri
2017-04-25 18:29   ` Stefan Berger
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 5/9] tmp backend: Add new api to read backend TpmInfo Amarnath Valluri
2017-04-25 18:51   ` Stefan Berger
2017-04-25 18:59     ` Eric Blake
2017-05-02  7:17       ` Amarnath Valluri
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 6/9] tpm-backend: Remove unneeded destroy() method from TpmDriverOps interface Amarnath Valluri
2017-04-25 18:59   ` Stefan Berger
2017-05-02  7:42     ` Amarnath Valluri
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 7/9] tpm-backend: Move realloc_buffer() implementation to base class Amarnath Valluri
2017-04-25 19:01   ` Stefan Berger
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 8/9] tpm-passthrough: move reusable code to utils Amarnath Valluri
2017-04-25 19:09   ` Stefan Berger
2017-04-07 14:30 ` [Qemu-devel] [PATCH v2 9/9] tpm: Added support for TPM emulator Amarnath Valluri
2017-04-07 14:41   ` Daniel P. Berrange [this message]
2017-04-07 15:11     ` Marc-André Lureau
2017-04-10  7:34       ` Amarnath Valluri
2017-04-10  9:54         ` Marc-André Lureau
2017-04-10 10:07           ` Patrick Ohly
2017-04-10 16:14             ` Stefan Berger
2017-04-10 21:11               ` Stefan Berger
2017-04-10  7:08     ` Amarnath Valluri
2017-04-10  8:31       ` Daniel P. Berrange
2017-04-10 16:15       ` Stefan Berger
2017-04-25 19:35   ` Stefan Berger
2017-04-12 23:52 ` [Qemu-devel] [PATCH v2 0/9] Provide support for the software TPM no-reply

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=20170407144100.GL26896@redhat.com \
    --to=berrange@redhat.com \
    --cc=amarnath.valluri@intel.com \
    --cc=marcandre.lureau@gmail.com \
    --cc=patrick.ohly@intel.com \
    --cc=qemu-devel@nongnu.org \
    --cc=stefanb@linux.vnet.ibm.com \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.