[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [PULL v2 2/5] hw/xen: Expose handle_bufioreq in xen_register_ioreq
On Thu, 3 Oct 2024 at 19:57, Edgar E. Iglesias <edgar.iglesias@xxxxxxxxx> wrote: > > From: "Edgar E. Iglesias" <edgar.iglesias@xxxxxxx> > > Expose handle_bufioreq in xen_register_ioreq(). > This is to allow machines to enable or disable buffered ioreqs. > > No functional change since all callers still set it to > HVM_IOREQSRV_BUFIOREQ_ATOMIC. > > Reviewed-by: Stefano Stabellini <sstabellini@xxxxxxxxxx> > Signed-off-by: Edgar E. Iglesias <edgar.iglesias@xxxxxxx> Hi; Coverity has noticed a problem (CID 1563383) with this change: > diff --git a/hw/xen/xen-hvm-common.c b/hw/xen/xen-hvm-common.c > index 3a9d6f981b..7d2b72853b 100644 > --- a/hw/xen/xen-hvm-common.c > +++ b/hw/xen/xen-hvm-common.c > @@ -667,6 +667,8 @@ static int xen_map_ioreq_server(XenIOState *state) > xen_pfn_t ioreq_pfn; > xen_pfn_t bufioreq_pfn; > evtchn_port_t bufioreq_evtchn; In this function bufioreq_evtchn is declared uninitialized... > + unsigned long num_frames = 1; > + unsigned long frame = 1; > int rc; > > /* > @@ -675,59 +677,78 @@ static int xen_map_ioreq_server(XenIOState *state) > */ > QEMU_BUILD_BUG_ON(XENMEM_resource_ioreq_server_frame_bufioreq != 0); > QEMU_BUILD_BUG_ON(XENMEM_resource_ioreq_server_frame_ioreq(0) != 1); > + > + if (state->has_bufioreq) { > + frame = 0; > + num_frames = 2; > + } > state->fres = xenforeignmemory_map_resource(xen_fmem, xen_domid, > XENMEM_resource_ioreq_server, > - state->ioservid, 0, 2, > + state->ioservid, > + frame, num_frames, > &addr, > PROT_READ | PROT_WRITE, 0); > if (state->fres != NULL) { > trace_xen_map_resource_ioreq(state->ioservid, addr); > - state->buffered_io_page = addr; > - state->shared_page = addr + XC_PAGE_SIZE; > + state->shared_page = addr; > + if (state->has_bufioreq) { > + state->buffered_io_page = addr; > + state->shared_page = addr + XC_PAGE_SIZE; > + } > } else if (errno != EOPNOTSUPP) { > error_report("failed to map ioreq server resources: error %d > handle=%p", > errno, xen_xc); > return -1; > } > > - rc = xen_get_ioreq_server_info(xen_domid, state->ioservid, > - (state->shared_page == NULL) ? > - &ioreq_pfn : NULL, > - (state->buffered_io_page == NULL) ? > - &bufioreq_pfn : NULL, > - &bufioreq_evtchn); ...which was OK prior to this change, because (ignoring the early-exit case) we would always pass through this function call, which initializes bufioreq_evtchn... > - if (rc < 0) { > - error_report("failed to get ioreq server info: error %d handle=%p", > - errno, xen_xc); > - return rc; > - } > + /* > + * If we fail to map the shared page with xenforeignmemory_map_resource() > + * or if we're using buffered ioreqs, we need xen_get_ioreq_server_info() > + * to provide the the addresses to map the shared page and/or to get the > + * event-channel port for buffered ioreqs. > + */ > + if (state->shared_page == NULL || state->has_bufioreq) { > + rc = xen_get_ioreq_server_info(xen_domid, state->ioservid, > + (state->shared_page == NULL) ? > + &ioreq_pfn : NULL, > + (state->has_bufioreq && > + state->buffered_io_page == NULL) ? > + &bufioreq_pfn : NULL, > + &bufioreq_evtchn); ...but now the initialization has moved inside an if() so it only happens under certain conditions... > + if (rc < 0) { > + error_report("failed to get ioreq server info: error %d > handle=%p", > + errno, xen_xc); > + return rc; > + } > > - if (state->shared_page == NULL) { > - trace_xen_map_ioreq_server_shared_page(ioreq_pfn); > + if (state->shared_page == NULL) { > + trace_xen_map_ioreq_server_shared_page(ioreq_pfn); > > - state->shared_page = xenforeignmemory_map(xen_fmem, xen_domid, > - PROT_READ | PROT_WRITE, > - 1, &ioreq_pfn, NULL); > + state->shared_page = xenforeignmemory_map(xen_fmem, xen_domid, > + PROT_READ | PROT_WRITE, > + 1, &ioreq_pfn, NULL); > + } > if (state->shared_page == NULL) { > error_report("map shared IO page returned error %d handle=%p", > errno, xen_xc); > } > - } > > - if (state->buffered_io_page == NULL) { > - trace_xen_map_ioreq_server_buffered_io_page(bufioreq_pfn); > + if (state->has_bufioreq && state->buffered_io_page == NULL) { > + trace_xen_map_ioreq_server_buffered_io_page(bufioreq_pfn); > > - state->buffered_io_page = xenforeignmemory_map(xen_fmem, xen_domid, > - PROT_READ | > PROT_WRITE, > - 1, &bufioreq_pfn, > - NULL); > - if (state->buffered_io_page == NULL) { > - error_report("map buffered IO page returned error %d", errno); > - return -1; > + state->buffered_io_page = xenforeignmemory_map(xen_fmem, > xen_domid, > + PROT_READ | > PROT_WRITE, > + 1, &bufioreq_pfn, > + NULL); > + if (state->buffered_io_page == NULL) { > + error_report("map buffered IO page returned error %d", > errno); > + return -1; > + } > } > } > > - if (state->shared_page == NULL || state->buffered_io_page == NULL) { > + if (state->shared_page == NULL || > + (state->has_bufioreq && state->buffered_io_page == NULL)) { > return -1; > } ...and the tail end of the function has not been modified, so (not visible in this diff context) when we do: trace_xen_map_ioreq_server_buffered_io_evtchn(bufioreq_evtchn); state->bufioreq_remote_port = bufioreq_evtchn; return 0; we may be using it uninitialized (in the trace statement and when assigning it to state->bufioreq_remote_port). Could you have a look at this and send a fix, please? thanks -- PMM
|
Lists.xenproject.org is hosted with RackSpace, monitoring our |