On 09/26/2011 03:09 PM, Michael S. Tsirkin wrote:
On Mon, Sep 26, 2011 at 12:35:11PM -0400, Stefan Berger wrote:
+static int tpm_tis_init(ISADevice *dev)
+{
+ TPMState *s = DO_UPCAST(TPMState, busdev, dev);
+ TPMTISState *tis =&s->s.tis;
+ int iomemtype, rc;
+
+ qemu_mutex_init(&s->state_lock);
+ qemu_cond_init(&s->from_tpm_cond);
+ qemu_cond_init(&s->to_tpm_cond);
+
+ s->be_driver = qemu_find_tpm(s->backend);
+ if (!s->be_driver) {
+ fprintf(stderr,
+ "tpm_tis: backend driver with id %s could not be found.n\n",
+ s->backend);
error_report?
Fixed.
+ return -1;
+ }
+
+ if (s->be_driver->ops->init(s->be_driver, s, tpm_tis_receive_cb)) {
+ goto err_exit;
+ }
+
+ isa_init_irq(dev,&tis->irq, tis->irq_num);
+
+ iomemtype = cpu_register_io_memory(tpm_tis_readfn, tpm_tis_writefn, s,
+ DEVICE_LITTLE_ENDIAN);
A bit of a strange name for a var - it's not a type, is it?
+ cpu_register_physical_memory(TPM_TIS_ADDR_BASE,
+ 0x1000 * TPM_TIS_NUM_LOCALITIES,
+ iomemtype);
I think memory_region_init_io are supposed to be used for new devices
from now on... Right, Avi?
I fixed that now also to use
memory_region_init_io(&s->mmio, &tpm_tis_memory_ops, s, "tpm-tis-mmio",
0x1000 * TPM_TIS_NUM_LOCALITIES);
memory_region_add_subregion(get_system_memory(), TPM_TIS_ADDR_BASE,
&s->mmio);
+
+ rc = tpm_tis_do_startup_tpm(s);
+ if (rc != 0) {
+ goto err_exit;
+ }
+
+ return 0;
+
+ err_exit:
Missing cleanup?
True. Fixed.
I had previously looked around in other devices how they were doing it
and had often seen them not cleaning up, either.
Stefan
