Could you give more explanation about pci posting flush? When (and why) do we need it? In an old posting (http://lkml.org/lkml/2003/5/8/278) said pci posting flush is unnecessary - is it correct? Thanks. > >> + } >> + >> + if (req != IOPMU_QUEUE_EMPTY) { > + writel(req, &iop->outbound_queue); > >does this need a PCI posting flush? > >> + >> +static int iop_send_sync_msg(struct hptiop_hba * hba, u32 msg, u32 >> +millisec) { >> + u32 i; >> + >> + hba->msg_done = 0; >> + >> + writel(msg, &hba->iop->inbound_msgaddr0); >> + >> + for (i = 0; i < millisec; i++) { >> + __iop_intr(hba); >> + if (hba->msg_done) >> + break; >> + mdelay(1); >> + } >> >>and here, but here you're very clearly missing a pci posting flush >> >> + else >> + arg->result = HPT_IOCTL_RESULT_FAILED; >> + >> + arg->done(arg); >> + writel(tag, &hba->iop->outbound_queue); } > >this looks like it needs a posting flush again > >> + memcpy(req->cdb, scp->cmnd, sizeof(req->cdb)); >> + >> + writel(IOPMU_QUEUE_ADDR_HOST_BIT | _req->req_shifted_phy, >> + &hba->iop->inbound_queue); > >this needs pci posting flush > >> + >> +static int hptiop_reset_hba(struct hptiop_hba * hba) { if >> +(atomic_xchg(&hba->resetting, 1) == 0) { >> + atomic_inc(&hba->reset_count); >> + writel(IOPMU_INBOUND_MSG0_RESET, >> + &hba->iop->outbound_msgaddr0); >> + } > >this needs a pci posting flush - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html