Re: [PATCH] xl create: endless loop

[Christoph Egger <Christoph.Egger@amd.com>]

On Thursday 21 October 2010 16:11:55 Stefano Stabellini wrote:
> On Tue, 19 Oct 2010, Christoph Egger wrote:
> > On Monday 18 October 2010 14:50:35 Christoph Egger wrote:
> > > Hi!
> > >
> > > I cannot start a guest with 'xl create' due to an endless loop in
> > > libxl.c, function libxl__get_free_memory_slack():
> > >
> > > There is this code snippet:
> > >
> > > retry:
> > >     free_mem_slack_s = libxl__xs_read(gc, XBT_NULL,
> > > free_mem_slack_path); if (!free_mem_slack_s) {
> > >         rc = libxl__fill_dom0_memory_info(gc, &target_memkb);
> > >         if (rc < 0)
> > >             return rc;
> > >         goto retry;
> > >     } else {
> > >
> > >
> > > libxl__xs_read() returns 0 and libxl__fill_dom0_memory_info() also
> > > returns 0. So there's a loop of retries.
> >
> > Attached patch fixes the endless loop.
> >
> > Signed-off-by: Christoph Egger <Christoph.Egger@amd.com>
>
> thanks for the patch but this cannot possibly be the right fix:
>
>
> diff -r 9d8d6b93114e tools/libxl/libxl.c
> --- a/tools/libxl/libxl.c	Mon Oct 18 15:24:39 2010 +0200
> +++ b/tools/libxl/libxl.c	Tue Oct 19 10:29:00 2010 +0200
> @@ -2836,7 +2836,7 @@ retry:
>      free_mem_slack_s = libxl__xs_read(gc, XBT_NULL, free_mem_slack_path);
>      if (!free_mem_slack_s) {
>          rc = libxl__fill_dom0_memory_info(gc, &target_memkb);
> -        if (rc < 0)
> +        if (rc <= 0)
>              return rc;
>          goto retry;
>      } else {
>
> the idea is that libxl__fill_dom0_memory_info should fill the missing
> informations in xenstore so that we can go ahead and try to read them
> again and the second time should be successful.
> Libxl__fill_dom0_memory_info returns 0 on success, so it is correct to
> goto retry in that case.
>
> The bug must be in libxl__fill_dom0_memory_info that doesn't return
> error when it should.
> Does the appended patch works for you?
>
> ---
>
> diff -r 00b92112b055 tools/libxl/libxl.c
> --- a/tools/libxl/libxl.c	Wed Oct 20 17:26:51 2010 +0100
> +++ b/tools/libxl/libxl.c	Thu Oct 21 15:08:29 2010 +0100
> @@ -2793,11 +2793,11 @@ retry_transaction:
>
>      rc = libxl_domain_info(ctx, &info, 0);
>      if (rc < 0)
> -        return rc;
> +        goto out;
>
>      rc = libxl_get_physinfo(ctx, &physinfo);
>      if (rc < 0)
> -        return rc;
> +        goto out;
>
>      libxl__xs_write(gc, t, target_path, "%"PRIu32,
>              (uint32_t) info.current_memkb);
> @@ -2816,9 +2816,12 @@ retry_transaction:
>      rc = 0;
>
>  out:
> -    if (!xs_transaction_end(ctx->xsh, t, 0))
> +    if (!xs_transaction_end(ctx->xsh, t, 0)) {
>          if (errno == EAGAIN)
>              goto retry_transaction;
> +        else
> +            rc = ERROR_FAIL;
> +    }
>
>
>      return rc;


No, this patch has no effect for me.
In libxl__fill_dom0_memory_info(), the code path goes that way:

    t = xs_transaction_start(ctx->xsh);

    target = libxl__xs_read(gc, t, target_path);
    if (target) {    <-- target contains "5"
        *target_memkb = strtoul(target, &endptr, 10);
        if (*endptr != '\0') {     <-- *endptr contains '\0'
            LIBXL__LOG_ERRNO(ctx, LIBXL__LOG_ERROR,
                    "invalid memory target %s from %s\n", target, 
target_path);
            rc = ERROR_FAIL;
            goto out;
        }
        rc = 0;
        goto out;    <-- take this jump with rc being 0
    }


Christoph



-- 
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Einsteinring 24, 85609 Dornach b. Muenchen
Geschaeftsfuehrer: Alberto Bozzo, Andrew Bowd
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632