From dc8900e8c679af4ca1a8b27a93fc3929a7f74a94 Mon Sep 17 00:00:00 2001 From: chuck Date: Fri, 5 Apr 2019 16:54:20 +0000 Subject: [PATCH] bhyve: Fix NVMe data structure copy to guest bhyve's NVMe emulation was transferring Identify data back to the guest incorrectly causing memory corruptions. These corruptions resulted in core dumps and other system level errors in the guest. In their simplest form, NVMe Physical Region Page (PRP) values in commands indicate which physical pages to use for data transfer. The first PRP value is not required to be page aligned but does not cross a page boundary. The second PRP value must be page aligned, does not cross a page boundary, and need not be contiguous with PRP1. The code was copying Identify data past the end of PRP1. This happens to work if PRP1 and PRP2 are physically contiguous but will corrupt guest memory in unpredictable ways if they are not. Fix is to copy the Identify data back to the guest piecewise (i.e. for each PRP entry). Also fix a similarly wrong problem when copying back Log page data. Reviewed by: imp (mentor), araujo, jhb, rgrimes, bhyve Approved by: imp (mentor), bhyve (jhb) MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D19695 --- usr.sbin/bhyve/pci_nvme.c | 85 +++++++++++++++++++++++++++++++-------- 1 file changed, 68 insertions(+), 17 deletions(-) diff --git a/usr.sbin/bhyve/pci_nvme.c b/usr.sbin/bhyve/pci_nvme.c index 61d37d1f083..dbd9298f5a5 100644 --- a/usr.sbin/bhyve/pci_nvme.c +++ b/usr.sbin/bhyve/pci_nvme.c @@ -202,6 +202,9 @@ struct pci_nvme_softc { struct nvme_namespace_data nsdata; struct nvme_controller_data ctrldata; + struct nvme_error_information_entry err_log; + struct nvme_health_information_page health_log; + struct nvme_firmware_page fw_log; struct pci_nvme_blockstore nvstore; @@ -368,6 +371,15 @@ pci_nvme_init_nsdata(struct pci_nvme_softc *sc) nd->flbas = 0; } +static void +pci_nvme_init_logpages(struct pci_nvme_softc *sc) +{ + + memset(&sc->err_log, 0, sizeof(sc->err_log)); + memset(&sc->health_log, 0, sizeof(sc->health_log)); + memset(&sc->fw_log, 0, sizeof(sc->fw_log)); +} + static void pci_nvme_reset_locked(struct pci_nvme_softc *sc) { @@ -457,6 +469,47 @@ pci_nvme_init_controller(struct vmctx *ctx, struct pci_nvme_softc *sc) __func__, sc->regs.acq, sc->compl_queues[0].qbase)); } +static int +nvme_prp_memcpy(struct vmctx *ctx, uint64_t prp1, uint64_t prp2, uint8_t *src, + size_t len) +{ + uint8_t *dst; + size_t bytes; + + if (len > (8 * 1024)) { + return (-1); + } + + /* Copy from the start of prp1 to the end of the physical page */ + bytes = PAGE_SIZE - (prp1 & PAGE_MASK); + bytes = MIN(bytes, len); + + dst = vm_map_gpa(ctx, prp1, bytes); + if (dst == NULL) { + return (-1); + } + + memcpy(dst, src, bytes); + + src += bytes; + + len -= bytes; + if (len == 0) { + return (0); + } + + len = MIN(len, PAGE_SIZE); + + dst = vm_map_gpa(ctx, prp2, len); + if (dst == NULL) { + return (-1); + } + + memcpy(dst, src, len); + + return (0); +} + static int nvme_opc_delete_io_sq(struct pci_nvme_softc* sc, struct nvme_command* command, struct nvme_completion* compl) @@ -590,26 +643,24 @@ nvme_opc_get_log_page(struct pci_nvme_softc* sc, struct nvme_command* command, { uint32_t logsize = (1 + ((command->cdw10 >> 16) & 0xFFF)) * 2; uint8_t logpage = command->cdw10 & 0xFF; - void *data; DPRINTF(("%s log page %u len %u\r\n", __func__, logpage, logsize)); - if (logpage >= 1 && logpage <= 3) - data = vm_map_gpa(sc->nsc_pi->pi_vmctx, command->prp1, - PAGE_SIZE); - pci_nvme_status_genc(&compl->status, NVME_SC_SUCCESS); switch (logpage) { - case 0x01: /* Error information */ - memset(data, 0, logsize > PAGE_SIZE ? PAGE_SIZE : logsize); + case NVME_LOG_ERROR: + nvme_prp_memcpy(sc->nsc_pi->pi_vmctx, command->prp1, + command->prp2, (uint8_t *)&sc->err_log, logsize); break; - case 0x02: /* SMART/Health information */ + case NVME_LOG_HEALTH_INFORMATION: /* TODO: present some smart info */ - memset(data, 0, logsize > PAGE_SIZE ? PAGE_SIZE : logsize); + nvme_prp_memcpy(sc->nsc_pi->pi_vmctx, command->prp1, + command->prp2, (uint8_t *)&sc->health_log, logsize); break; - case 0x03: /* Firmware slot information */ - memset(data, 0, logsize > PAGE_SIZE ? PAGE_SIZE : logsize); + case NVME_LOG_FIRMWARE_SLOT: + nvme_prp_memcpy(sc->nsc_pi->pi_vmctx, command->prp1, + command->prp2, (uint8_t *)&sc->fw_log, logsize); break; default: WPRINTF(("%s get log page %x command not supported\r\n", @@ -633,14 +684,13 @@ nvme_opc_identify(struct pci_nvme_softc* sc, struct nvme_command* command, switch (command->cdw10 & 0xFF) { case 0x00: /* return Identify Namespace data structure */ - dest = vm_map_gpa(sc->nsc_pi->pi_vmctx, command->prp1, - sizeof(sc->nsdata)); - memcpy(dest, &sc->nsdata, sizeof(sc->nsdata)); + nvme_prp_memcpy(sc->nsc_pi->pi_vmctx, command->prp1, + command->prp2, (uint8_t *)&sc->nsdata, sizeof(sc->nsdata)); break; case 0x01: /* return Identify Controller data structure */ - dest = vm_map_gpa(sc->nsc_pi->pi_vmctx, command->prp1, - sizeof(sc->ctrldata)); - memcpy(dest, &sc->ctrldata, sizeof(sc->ctrldata)); + nvme_prp_memcpy(sc->nsc_pi->pi_vmctx, command->prp1, + command->prp2, (uint8_t *)&sc->ctrldata, + sizeof(sc->ctrldata)); break; case 0x02: /* list of 1024 active NSIDs > CDW1.NSID */ dest = vm_map_gpa(sc->nsc_pi->pi_vmctx, command->prp1, @@ -1881,6 +1931,7 @@ pci_nvme_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) pci_nvme_reset(sc); pci_nvme_init_ctrldata(sc); pci_nvme_init_nsdata(sc); + pci_nvme_init_logpages(sc); pci_lintr_request(pi); -- 2.45.0