Re: [Xen-devel] [PATCH 15/15] libxl: New event generation API

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Tue, 13 Dec 2011, Ian Campbell wrote:
> On Mon, 2011-12-12 at 19:24 +0000, Stefano Stabellini wrote:
> > diff --git a/tools/libxl/libxl.c b/tools/libxl/libxl.c
> > index cd613f7..94bdbba 100644
> > --- a/tools/libxl/libxl.c
> > +++ b/tools/libxl/libxl.c
> > @@ -63,6 +63,12 @@ int libxl_ctx_alloc(libxl_ctx **pctx, int version,
> >       * only as an initialiser, not as an expression. */
> >      memcpy(&ctx->lock, &mutex_value, sizeof(ctx->lock));
> >  
> > +    if(pthread_key_create(&ctx->tls_key, NULL) < 0) {
> > +        LIBXL__LOG_ERRNOVAL(ctx, LIBXL__LOG_ERROR, errno,
> > +            "cannot create tls key");
> > +        return ERROR_FAIL;
> 
> You need a corresponding pthread_key_delete in the free function.

good point

> > [...]
> 
> > @@ -65,17 +66,41 @@ int libxl__ptr_add(libxl__gc *gc, void *ptr)
> >      return 0;
> >  }
> >  
> > +libxl__gc *libxl__init_gc(libxl_ctx *ctx)
> > +{
> > +    libxl__gc *gc = (libxl__gc *) pthread_getspecific(ctx->tls_key);
> > +    if (gc == NULL) {
> > +        gc = (libxl__gc *) malloc(sizeof(libxl__gc));
> > +        if (gc == NULL)
> > +            return NULL;
> > +        gc->alloc_maxsize = 0;
> > +        gc->alloc_ptrs = 0;
> > +        gc->owner = ctx;
> > +        gc->nested = 1;
> > +        pthread_setspecific(ctx->tls_key, gc);
> 
> pthread_setspecific can fail. 

right, I'll handle that

> > [...]
> 
> >  void libxl__free_all(libxl__gc *gc)
> >  {
> >      void *ptr;
> >      int i;
> >  
> > +    gc->nested--;
> > +    if (gc->nested > 0)
> > +        return;
> > +
> >      for (i = 0; i < gc->alloc_maxsize; i++) {
> >          ptr = gc->alloc_ptrs[i];
> >          gc->alloc_ptrs[i] = NULL;
> >          free(ptr);
> >      }
> >      free(gc->alloc_ptrs);
> > +    pthread_setspecific(CTX->tls_key, NULL);
> 
> As above this can also fail. I think this one might be a bit harder to
> deal with in general.

this pthread_setspecific call, with the NULL paramter, can only fail if
the key is invalid, so I think that we don't need to handle that error.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel