lkml.org 
[lkml]   [2019]   [Jul]   [15]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    From
    Date
    SubjectRe: [PATCH v9 02/18] kunit: test: add test resource management API
    On Mon, Jul 15, 2019 at 1:24 PM Stephen Boyd <sboyd@kernel.org> wrote:
    >
    > Quoting Brendan Higgins (2019-07-12 01:17:28)
    > > diff --git a/kunit/test.c b/kunit/test.c
    > > index 571e4c65deb5c..f165c9d8e10b0 100644
    > > --- a/kunit/test.c
    > > +++ b/kunit/test.c
    > > @@ -171,6 +175,96 @@ int kunit_run_tests(struct kunit_suite *suite)
    > > return 0;
    > > }
    > >
    > > +struct kunit_resource *kunit_alloc_resource(struct kunit *test,
    > > + kunit_resource_init_t init,
    > > + kunit_resource_free_t free,
    > > + void *context)
    > > +{
    > > + struct kunit_resource *res;
    > > + int ret;
    > > +
    > > + res = kzalloc(sizeof(*res), GFP_KERNEL);
    >
    > This uses GFP_KERNEL.
    >
    > > + if (!res)
    > > + return NULL;
    > > +
    > > + ret = init(res, context);
    > > + if (ret)
    > > + return NULL;
    > > +
    > > + res->free = free;
    > > + mutex_lock(&test->lock);
    >
    > And this can sleep.
    >
    > > + list_add_tail(&res->node, &test->resources);
    > > + mutex_unlock(&test->lock);
    > > +
    > > + return res;
    > > +}
    > > +
    > > +void kunit_free_resource(struct kunit *test, struct kunit_resource *res)
    >
    > Should probably add a note that we assume the test lock is held here, or
    > even add a lockdep_assert_held(&test->lock) into the function to
    > document that and assert it at the same time.

    Seems reasonable.

    > > +{
    > > + res->free(res);
    > > + list_del(&res->node);
    > > + kfree(res);
    > > +}
    > > +
    > > +struct kunit_kmalloc_params {
    > > + size_t size;
    > > + gfp_t gfp;
    > > +};
    > > +
    > > +static int kunit_kmalloc_init(struct kunit_resource *res, void *context)
    > > +{
    > > + struct kunit_kmalloc_params *params = context;
    > > +
    > > + res->allocation = kmalloc(params->size, params->gfp);
    > > + if (!res->allocation)
    > > + return -ENOMEM;
    > > +
    > > + return 0;
    > > +}
    > > +
    > > +static void kunit_kmalloc_free(struct kunit_resource *res)
    > > +{
    > > + kfree(res->allocation);
    > > +}
    > > +
    > > +void *kunit_kmalloc(struct kunit *test, size_t size, gfp_t gfp)
    > > +{
    > > + struct kunit_kmalloc_params params;
    > > + struct kunit_resource *res;
    > > +
    > > + params.size = size;
    > > + params.gfp = gfp;
    > > +
    > > + res = kunit_alloc_resource(test,
    >
    > This calls that sleeping function above...
    >
    > > + kunit_kmalloc_init,
    > > + kunit_kmalloc_free,
    > > + &params);
    >
    > but this passes a GFP flags parameter through to the
    > kunit_kmalloc_init() function. How is this going to work if some code
    > uses GFP_ATOMIC, but then we try to allocate and sleep in
    > kunit_alloc_resource() with GFP_KERNEL?

    Yeah, that's an inconsistency. I need to fix that.

    > One solution would be to piggyback on all the existing devres allocation
    > logic we already have and make each struct kunit a device that we pass
    > into the devres functions. A far simpler solution would be to just
    > copy/paste what devres does and use a spinlock and an allocation
    > function that takes GFP flags.

    Yeah, that's what I did originally, but I thought from the discussion
    on patch 01 that you thought a spinlock was overkill for struct kunit.
    I take it you only meant in that initial patch?

    > > +
    > > + if (res)
    > > + return res->allocation;
    > > +
    > > + return NULL;
    > > +}

    Cheers

    \
     
     \ /
      Last update: 2019-07-15 22:31    [W:4.033 / U:0.436 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site