SeaBIOS October 2017

seabios@seabios.org

17 participants
38 discussions

答复: Welcome to the "SeaBIOS" mailing list (Digest mode)
by 李勇 21 Oct '17

21 Oct '17

-----邮件原件----- 发件人: SeaBIOS [mailto:seabios-bounces@seabios.org] 代表 seabios-request(a)seabios.org 发送时间: 2017年10月20日 7:35 收件人: edwardli(a)dl.cn 主题: Welcome to the "SeaBIOS" mailing list (Digest mode) Welcome to the SeaBIOS(a)seabios.org mailing list! To post to this list, send your message to: seabios(a)seabios.org General information about the mailing list is at: https://mail.coreboot.org/mailman/listinfo/seabios If you ever want to unsubscribe or change your options (eg, switch to or from digest mode, change your password, etc.), visit your subscription page at: https://mail.coreboot.org/mailman/options/seabios/edwardli%40dl.cn You can also make such adjustments via email by sending a message to: SeaBIOS-request(a)seabios.org with the word `help' in the subject or body (don't include the quotes), and you will get back a message with instructions. You must know your password to change your options (including changing the password, itself) or to unsubscribe without confirmation. It is: nLixZhot Normally, Mailman will remind you of your seabios.org mailing list passwords once every month, although you can disable this if you prefer. This reminder will also include instructions on how to unsubscribe or change your account options. There is also a button on your options page that will email your current password to you.

1 0

Failure to detect high-capacity SD card
by Matt DeVillier 20 Oct '17

20 Oct '17

I have a user running one of my SeaBIOS builds on a Asus Chromebook C302sa (google/cave) who is having trouble with high-capacity SD cards - they aren't detected at all due to timing out during init. I've tried adjusting the various timeouts, and the only change is longer log files. The SD card in question is a Samsung MB-MC128GA U3 Micro SDXC 128GB EVO Plus. Trimmed log below. The function start indicators were inserted by me as part of debugging. thanks in advance, Matt SeaBIOS (version rel-1.10.0-81-g0fff851-MrChromebox-2017.10.19) /7a9fe000\ Start thread |7a9fe000| Searching bootorder for: /pci@i0cf8/*@1e,6 |7a9fe000| sdcard_pci_setup() start |7a9fe000| sdcard_controller_setup() start |7a9fe000| sdhci@0xd1138000 ver=1002 cap=7568c881 807 |7a9fe000| sdcard_card_setup() start |7a9fe000| sdcard_set_frequency 200 400 fa00 |7a9fe000| sdcard_pio cmd 0 0 1ff0000 |7a9fe000| sdcard cmd 0 response 0 0 0 0 |7a9fe000| sdcard_pio cmd 81a 1aa 1ff0000 |7a9fe000| sdcard cmd 81a response 1aa 0 0 0 |7a9fe000| sdcard_pio cmd 371a 0 1ff0000 |7a9fe000| sdcard cmd 371a response 120 0 0 0 |7a9fe000| sdcard_pio cmd 2902 0 1ff0000 |7a9fe000| sdcard cmd 2902 response ff8000 0 0 0 |7a9fe000| sdcard_pio cmd 371a 0 1ff0000 |7a9fe000| sdcard cmd 371a response 120 0 0 0 |7a9fe000| sdcard_pio cmd 2902 40100000 1ff0000 |7a9fe000| sdcard cmd 2902 response ff8000 0 0 0 |7a9fe000| sdcard_pio cmd 371a 0 1ff0000 |7a9fe000| sdcard cmd 371a response 120 0 0 0 |7a9fe000| sdcard_pio cmd 2902 40100000 1ff0000 |7a9fe000| sdcard cmd 2902 response ff8000 0 0 0 <snip ~1s of 371a/2902 sequence repeating> |7a9fe000| WARNING - Timeout at sdcard_card_setup:432! \7a9fe000/ End thread All threads complete.

1 0

tpm: TPM 2 support
by 王杰 19 Oct '17

19 Oct '17

Hello, I am using the seabios-1.10.0 with qemu2.6, I know seabios-1.10.0 supports TPM2.0, but unfortunately it doesn’t work for me. I have tested CentOS 7.2 and Ubuntu 14.04, it all said: “ima: NO TPM chip found, activating TPM-bypass!” I doubt it’s a problem about TPM driver, so I download the Ubuntu kernel source code, and I find the interface as follows: --- TPM Hardware Support -*- TPM Interface Specification 1.2 Interface / TPM 2.0 FIFO Interface <M> TPM Interface Specification 1.2 Interface (I2C - Atmel) <M> TPM Interface Specification 1.2 Interface (I2C -Infineon) <M> TPM Interface Specification 1.2 Interface (I2C - Muvoton) <M> National Semiconductor TPM Interface <M> Atmel TPM Interface <M> Infineon Technologies TPM Interface <M> XEN TPM Interface < > TPM 2.0 CRB Interface < > STMicroelectronics TPM Interface Specification 1.2 Interface --- Should I need open the TPM 2.0 CRB Interface? I really hope you could give me some advice. Best Reagrds, Jie Wang

2 1

[PATCH] nvme: Use the Maximum Queue Entries Supported (MQES) to initialize I/O queues
by Filippo Sironi 15 Oct '17

15 Oct '17

... and cap the length to 256 to avoid oom-ing. Signed-off-by: Filippo Sironi <sironi(a)amazon.de> --- src/hw/nvme.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/src/hw/nvme.c b/src/hw/nvme.c index 946487f4fd43..19a4e7a48d3c 100644 --- a/src/hw/nvme.c +++ b/src/hw/nvme.c @@ -20,6 +20,8 @@ #include "nvme.h" #include "nvme-int.h" +#define min(a, b) ({ typeof(a) _a = a; typeof(b) _b = b; _a < _b ? _a : _b; }) + static void * zalloc_page_aligned(struct zone_s *zone, u32 size) { @@ -318,8 +320,10 @@ nvme_create_io_cq(struct nvme_ctrl *ctrl, struct nvme_cq *cq, u16 q_idx) { int rc; struct nvme_sqe *cmd_create_cq; + u16 length; - rc = nvme_init_cq(ctrl, cq, q_idx, NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + length = min(1 + (ctrl->reg->cap & 0xffff), NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + rc = nvme_init_cq(ctrl, cq, q_idx, length); if (rc) { goto err; } @@ -359,8 +363,10 @@ nvme_create_io_sq(struct nvme_ctrl *ctrl, struct nvme_sq *sq, u16 q_idx, struct { int rc; struct nvme_sqe *cmd_create_sq; + u16 length; - rc = nvme_init_sq(ctrl, sq, q_idx, NVME_PAGE_SIZE / sizeof(struct nvme_cqe), cq); + length = min(1 + (ctrl->reg->cap & 0xffff), NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + rc = nvme_init_sq(ctrl, sq, q_idx, length, cq); if (rc) { goto err; } -- 2.7.4

3 6

GCC 7.1 warning in `tcgbios.c`: `'hleo.eventnumber' may be used uninitialized in this function`
by Paul Menzel 12 Oct '17

12 Oct '17

Dear SeaBIOS folks, GCC 7.1 warns about a maybe uninitialized struct member. ``` src/tcgbios.c: In function 'tpm_interrupt_handler32': src/tcgbios.c:1232:30: warning: 'hleo.eventnumber' may be used uninitialized in this function [-Wmaybe-uninitialized] hleeo->eventnumber = hleo.eventnumber; ~~~~^~~~~~~~~~~~ Compiling whole program out/ccode32flat.o In file included from out/ccode32flat.o.tmp.c:52:0: ./src/tcgbios.c: In function 'tpm_interrupt_handler32': ./src/tcgbios.c:1232:30: warning: 'hleo.eventnumber' may be used uninitialized in this function [-Wmaybe-uninitialized] hleeo->eventnumber = hleo.eventnumber; ~~~~^~~~~~~~~~~~ ``` It looks valid, as in the beginning the struct hleo is only declared. ``` 1189 struct hleo hleo; ``` Kind regards, Paul

4 6

Graphics card driving cannot be loaded
by 柘龙义 11 Oct '17

11 Oct '17

When i running my vitural machine, it is normal except video And my mouse and key can be pressed , But my graphics is unnormal picture is very large in screen, Then i reboot VM , it is well. Why i get this and how to resolve it,Can you help me. thank you . My name is Lying.

1 0

Re: [SeaBIOS] How to boot USB 3.0 devices with SeaBIOS
by diffusae 10 Oct '17

10 Oct '17

Hello, thanks for help. On 03.10.2017 18:15, Kevin O'Connor wrote: > Can you recompile SeaBIOS with the patch below applied? I don't have > a good way to test this, but I'm hopeful it will help detect the USB2 > disconnect and avoid the controller hang. I've applied the patch. But maybe I need some help with this. Coreboot will always overwrite these changes. error: Your local changes to the following files would be overwritten by checkout: src/hw/usb-xhci.c Please, commit your changes or stash them before you can switch branches. Aborting I've tried stash, but that really won't work. I went back to version rel-1.10.0-70-gf3d2a15, applied the patch and copy it to the master branch. Don't know, if I've done a "checkout -- <file>" or something else, but after all, I got it working and can compile with the patch. But I don't know how. :-) Now, I am on version: rel-1.10.0-72-gf703604-dirty-20171008_051220-x220. The controller doesn't hang any more and the device is detected at boot. I'm not sure, if it's only USB2. Because of the verbose logging, the cbmem console log is truncated. I've attached the debug files. Hopefully you can commit you're changes. If you need further testing, please send a message. Best regards, Reiner --

3 3

Jenkins build is back to normal : seabios #270
by Jenkins Build Host 06 Oct '17

06 Oct '17

See <http://jenkins.xeni.kraxel.org:8080/job/seabios/270/display/redirect?page=c…>

1 0

[PATCH] xhci: Build TRBs directly in xhci_trb_queue()
by Kevin O'Connor 06 Oct '17

06 Oct '17

Use the logic for building a 'struct xhci_trb' that was in xhci_xfer_queue() up so that command and ring TRBs can also use that functionality. This eliminates the need to manually generate the xhci_trb struct from those code paths. Signed-off-by: Kevin O'Connor <kevin(a)koconnor.net> --- src/hw/usb-xhci.c | 162 +++++++++++++++++++----------------------------------- 1 file changed, 57 insertions(+), 105 deletions(-) diff --git a/src/hw/usb-xhci.c b/src/hw/usb-xhci.c index 50b3b86..a92a162 100644 --- a/src/hw/usb-xhci.c +++ b/src/hw/usb-xhci.c @@ -641,13 +641,16 @@ xhci_setup(void) * End point communication ****************************************************************/ +// Signal the hardware to process events on a TRB ring static void xhci_doorbell(struct usb_xhci_s *xhci, u32 slotid, u32 value) { + dprintf(5, "%s: slotid %d, epid %d\n", __func__, slotid, value); struct xhci_db *db = xhci->db; void *addr = &db[slotid].doorbell; writel(addr, value); } +// Dequeue events on the XHCI command ring generated by the hardware static void xhci_process_events(struct usb_xhci_s *xhci) { struct xhci_ring *evts = xhci->evts; @@ -712,6 +715,7 @@ static void xhci_process_events(struct usb_xhci_s *xhci) } } +// Check if a ring has any pending TRBs static int xhci_ring_busy(struct xhci_ring *ring) { u32 eidx = ring->eidx; @@ -719,6 +723,7 @@ static int xhci_ring_busy(struct xhci_ring *ring) return (eidx != nidx); } +// Wait for a ring to empty (all TRBs processed by hardware) static int xhci_event_wait(struct usb_xhci_s *xhci, struct xhci_ring *ring, u32 timeout) @@ -739,69 +744,54 @@ static int xhci_event_wait(struct usb_xhci_s *xhci, } } -static void xhci_trb_queue(struct xhci_ring *ring, - struct xhci_trb *trb) +// Add a TRB to the given ring +static void xhci_trb_fill(struct xhci_ring *ring + , void *data, u32 xferlen, u32 flags) { - u32 nidx = ring->nidx; - u32 cs = ring->cs; - struct xhci_trb *dst; - u32 control; - - if (nidx == XHCI_RING_ITEMS-1) { - dst = ring->ring + nidx; - control = (TR_LINK << 10); // trb type - control |= TRB_LK_TC; - control |= (cs ? TRB_C : 0); - dst->ptr_low = (u32)&ring[0]; + struct xhci_trb *dst = &ring->ring[ring->nidx]; + if (flags & TRB_TR_IDT) { + memcpy(&dst->ptr_low, data, xferlen); + } else { + dst->ptr_low = (u32)data; dst->ptr_high = 0; - dst->status = 0; - dst->control = control; - nidx = 0; - cs = cs ? 0 : 1; - ring->nidx = nidx; - ring->cs = cs; + } + dst->status = xferlen; + dst->control = flags | (ring->cs ? TRB_C : 0); +} +// Queue a TRB onto a ring, wrapping ring as needed +static void xhci_trb_queue(struct xhci_ring *ring, + void *data, u32 xferlen, u32 flags) +{ + if (ring->nidx >= ARRAY_SIZE(ring->ring) - 1) { + xhci_trb_fill(ring, ring->ring, 0, (TR_LINK << 10) | TRB_LK_TC); + ring->nidx = 0; + ring->cs ^= 1; dprintf(5, "%s: ring %p [linked]\n", __func__, ring); } - dst = ring->ring + nidx; - control = trb->control | (cs ? TRB_C : 0); - - dst->ptr_low = trb->ptr_low; - dst->ptr_high = trb->ptr_high; - dst->status = trb->status; - dst->control = control; - nidx++; - ring->nidx = nidx; - + xhci_trb_fill(ring, data, xferlen, flags); + ring->nidx++; dprintf(5, "%s: ring %p [nidx %d, len %d]\n", - __func__, ring, nidx, - trb->status & 0xffff); + __func__, ring, ring->nidx, xferlen); } -static int xhci_cmd_submit(struct usb_xhci_s *xhci, - struct xhci_trb *cmd) +// Submit a command to the xhci controller ring +static int xhci_cmd_submit(struct usb_xhci_s *xhci, struct xhci_inctx *inctx + , u32 flags) { - int rc; - mutex_lock(&xhci->cmds->lock); - xhci_trb_queue(xhci->cmds, cmd); + xhci_trb_queue(xhci->cmds, inctx, 0, flags); xhci_doorbell(xhci, 0, 0); - rc = xhci_event_wait(xhci, xhci->cmds, 1000); + int rc = xhci_event_wait(xhci, xhci->cmds, 1000); mutex_unlock(&xhci->cmds->lock); return rc; } static int xhci_cmd_enable_slot(struct usb_xhci_s *xhci) { - struct xhci_trb cmd = { - .ptr_low = 0, - .ptr_high = 0, - .status = 0, - .control = (CR_ENABLE_SLOT << 10) - }; dprintf(3, "%s:\n", __func__); - int cc = xhci_cmd_submit(xhci, &cmd); + int cc = xhci_cmd_submit(xhci, NULL, CR_ENABLE_SLOT << 10); if (cc != CC_SUCCESS) return -1; return (xhci->cmds->evt.control >> 24) & 0xff; @@ -809,55 +799,34 @@ static int xhci_cmd_enable_slot(struct usb_xhci_s *xhci) static int xhci_cmd_disable_slot(struct usb_xhci_s *xhci, u32 slotid) { - struct xhci_trb cmd = { - .ptr_low = 0, - .ptr_high = 0, - .status = 0, - .control = (slotid << 24) | (CR_DISABLE_SLOT << 10) - }; dprintf(3, "%s: slotid %d\n", __func__, slotid); - return xhci_cmd_submit(xhci, &cmd); + return xhci_cmd_submit(xhci, NULL, (CR_DISABLE_SLOT << 10) | (slotid << 24)); } static int xhci_cmd_address_device(struct usb_xhci_s *xhci, u32 slotid , struct xhci_inctx *inctx) { - struct xhci_trb cmd = { - .ptr_low = (u32)inctx, - .ptr_high = 0, - .status = 0, - .control = (slotid << 24) | (CR_ADDRESS_DEVICE << 10) - }; dprintf(3, "%s: slotid %d\n", __func__, slotid); - return xhci_cmd_submit(xhci, &cmd); + return xhci_cmd_submit(xhci, inctx + , (CR_ADDRESS_DEVICE << 10) | (slotid << 24)); } static int xhci_cmd_configure_endpoint(struct usb_xhci_s *xhci, u32 slotid , struct xhci_inctx *inctx) { - struct xhci_trb cmd = { - .ptr_low = (u32)inctx, - .ptr_high = 0, - .status = 0, - .control = (slotid << 24) | (CR_CONFIGURE_ENDPOINT << 10) - }; dprintf(3, "%s: slotid %d, add 0x%x, del 0x%x\n", __func__, slotid, inctx->add, inctx->del); - return xhci_cmd_submit(xhci, &cmd); + return xhci_cmd_submit(xhci, inctx + , (CR_CONFIGURE_ENDPOINT << 10) | (slotid << 24)); } static int xhci_cmd_evaluate_context(struct usb_xhci_s *xhci, u32 slotid , struct xhci_inctx *inctx) { - struct xhci_trb cmd = { - .ptr_low = (u32)inctx, - .ptr_high = 0, - .status = 0, - .control = (slotid << 24) | (CR_EVALUATE_CONTEXT << 10) - }; dprintf(3, "%s: slotid %d, add 0x%x, del 0x%x\n", __func__, slotid, inctx->add, inctx->del); - return xhci_cmd_submit(xhci, &cmd); + return xhci_cmd_submit(xhci, inctx + , (CR_EVALUATE_CONTEXT << 10) | (slotid << 24)); } static struct xhci_inctx * @@ -1092,37 +1061,29 @@ xhci_realloc_pipe(struct usbdevice_s *usbdev, struct usb_pipe *upipe return upipe; } -static void xhci_xfer_queue(struct xhci_pipe *pipe, - void *data, int datalen, u32 flags) -{ - struct xhci_trb trb; - memset(&trb, 0, sizeof(trb)); - if (flags & TRB_TR_IDT) - memcpy(&trb.ptr_low, data, datalen); - else - trb.ptr_low = (u32)data; - trb.status = datalen; - trb.control = flags; - xhci_trb_queue(&pipe->reqs, &trb); -} - -static void xhci_xfer_kick(struct xhci_pipe *pipe) +static void xhci_xfer_setup(struct xhci_pipe *pipe, int dir, void *cmd + , void *data, int datalen) { struct usb_xhci_s *xhci = container_of( pipe->pipe.cntl, struct usb_xhci_s, usb); - u32 slotid = pipe->slotid; - u32 epid = pipe->epid; - - dprintf(5, "%s: ring %p, slotid %d, epid %d\n", - __func__, &pipe->reqs, slotid, epid); - xhci_doorbell(xhci, slotid, epid); + xhci_trb_queue(&pipe->reqs, cmd, USB_CONTROL_SETUP_SIZE + , (TR_SETUP << 10) | TRB_TR_IDT + | ((datalen ? (dir ? 3 : 2) : 0) << 16)); + if (datalen) + xhci_trb_queue(&pipe->reqs, data, datalen, (TR_DATA << 10) + | ((dir ? 1 : 0) << 16)); + xhci_trb_queue(&pipe->reqs, NULL, 0, (TR_STATUS << 10) | TRB_TR_IOC + | ((dir ? 0 : 1) << 16)); + xhci_doorbell(xhci, pipe->slotid, pipe->epid); } static void xhci_xfer_normal(struct xhci_pipe *pipe, void *data, int datalen) { - xhci_xfer_queue(pipe, data, datalen, (TR_NORMAL << 10) | TRB_TR_IOC); - xhci_xfer_kick(pipe); + struct usb_xhci_s *xhci = container_of( + pipe->pipe.cntl, struct usb_xhci_s, usb); + xhci_trb_queue(&pipe->reqs, data, datalen, (TR_NORMAL << 10) | TRB_TR_IOC); + xhci_doorbell(xhci, pipe->slotid, pipe->epid); } int @@ -1140,16 +1101,7 @@ xhci_send_pipe(struct usb_pipe *p, int dir, const void *cmd if (req->bRequest == USB_REQ_SET_ADDRESS) // Set address command sent during xhci_alloc_pipe. return 0; - - xhci_xfer_queue(pipe, (void*)req, USB_CONTROL_SETUP_SIZE - , (TR_SETUP << 10) | TRB_TR_IDT - | ((datalen ? (dir ? 3 : 2) : 0) << 16)); - if (datalen) - xhci_xfer_queue(pipe, data, datalen, (TR_DATA << 10) - | ((dir ? 1 : 0) << 16)); - xhci_xfer_queue(pipe, NULL, 0, (TR_STATUS << 10) | TRB_TR_IOC - | ((dir ? 0 : 1) << 16)); - xhci_xfer_kick(pipe); + xhci_xfer_setup(pipe, dir, (void*)req, data, datalen); } else { xhci_xfer_normal(pipe, data, datalen); } -- 2.9.5

1 1

[PATCH] nvme: fix out of memory behavior
by Julian Stecklina 06 Oct '17

06 Oct '17

If the allocation of I/O queues ran out of memory, the code would fail to detect that and happily use these queues at address zero. For me this happens for systems with more than 7 NVMe controllers. Fix the out of memory handling to gracefully handle this case. Signed-off-by: Julian Stecklina <jsteckli(a)amazon.de> --- src/hw/nvme.c | 147 ++++++++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 111 insertions(+), 36 deletions(-) diff --git a/src/hw/nvme.c b/src/hw/nvme.c index 93c25f5..946487f 100644 --- a/src/hw/nvme.c +++ b/src/hw/nvme.c @@ -38,29 +38,43 @@ nvme_init_queue_common(struct nvme_ctrl *ctrl, struct nvme_queue *q, u16 q_idx, q->mask = length - 1; } -static void +static int nvme_init_sq(struct nvme_ctrl *ctrl, struct nvme_sq *sq, u16 q_idx, u16 length, struct nvme_cq *cq) { nvme_init_queue_common(ctrl, &sq->common, q_idx, length); sq->sqe = zalloc_page_aligned(&ZoneHigh, sizeof(*sq->sqe) * length); + + if (!sq->sqe) { + warn_noalloc(); + return -1; + } + dprintf(3, "sq %p q_idx %u sqe %p\n", sq, q_idx, sq->sqe); sq->cq = cq; sq->head = 0; sq->tail = 0; + + return 0; } -static void +static int nvme_init_cq(struct nvme_ctrl *ctrl, struct nvme_cq *cq, u16 q_idx, u16 length) { nvme_init_queue_common(ctrl, &cq->common, q_idx, length); cq->cqe = zalloc_page_aligned(&ZoneHigh, sizeof(*cq->cqe) * length); + if (!cq->cqe) { + warn_noalloc(); + return -1; + } cq->head = 0; /* All CQE phase bits are initialized to zero. This means initially we wait for the host controller to set these to 1. */ cq->phase = 1; + + return 0; } static int @@ -76,7 +90,6 @@ nvme_is_cqe_success(struct nvme_cqe const *cqe) return ((cqe->status >> 1) & 0xFF) == 0; } - static struct nvme_cqe nvme_error_cqe(void) { @@ -278,22 +291,44 @@ nvme_probe_ns(struct nvme_ctrl *ctrl, struct nvme_namespace *ns, u32 ns_id) dprintf(3, "%s", desc); boot_add_hd(&ns->drive, desc, bootprio_find_pci_device(ctrl->pci)); - free_buffer: +free_buffer: free (id); - } +} + + +/* Release memory allocated for a completion queue */ +static void +nvme_destroy_cq(struct nvme_cq *cq) +{ + free(cq->cqe); + cq->cqe = NULL; +} + +/* Release memory allocated for a submission queue */ +static void +nvme_destroy_sq(struct nvme_sq *sq) +{ + free(sq->sqe); + sq->sqe = NULL; +} /* Returns 0 on success. */ static int nvme_create_io_cq(struct nvme_ctrl *ctrl, struct nvme_cq *cq, u16 q_idx) { + int rc; struct nvme_sqe *cmd_create_cq; - nvme_init_cq(ctrl, cq, q_idx, NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + rc = nvme_init_cq(ctrl, cq, q_idx, NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + if (rc) { + goto err; + } + cmd_create_cq = nvme_get_next_sqe(&ctrl->admin_sq, NVME_SQE_OPC_ADMIN_CREATE_IO_CQ, NULL, cq->cqe); if (!cmd_create_cq) { - return -1; + goto err_destroy_cq; } cmd_create_cq->dword[10] = (cq->common.mask << 16) | (q_idx >> 1); @@ -307,24 +342,34 @@ nvme_create_io_cq(struct nvme_ctrl *ctrl, struct nvme_cq *cq, u16 q_idx) dprintf(2, "create io cq failed: %08x %08x %08x %08x\n", cqe.dword[0], cqe.dword[1], cqe.dword[2], cqe.dword[3]); - return -1; + goto err_destroy_cq; } return 0; + +err_destroy_cq: + nvme_destroy_cq(cq); +err: + return -1; } /* Returns 0 on success. */ static int nvme_create_io_sq(struct nvme_ctrl *ctrl, struct nvme_sq *sq, u16 q_idx, struct nvme_cq *cq) { + int rc; struct nvme_sqe *cmd_create_sq; - nvme_init_sq(ctrl, sq, q_idx, NVME_PAGE_SIZE / sizeof(struct nvme_cqe), cq); + rc = nvme_init_sq(ctrl, sq, q_idx, NVME_PAGE_SIZE / sizeof(struct nvme_cqe), cq); + if (rc) { + goto err; + } + cmd_create_sq = nvme_get_next_sqe(&ctrl->admin_sq, NVME_SQE_OPC_ADMIN_CREATE_IO_SQ, NULL, sq->sqe); if (!cmd_create_sq) { - return -1; + goto err_destroy_sq; } cmd_create_sq->dword[10] = (sq->common.mask << 16) | (q_idx >> 1); @@ -339,10 +384,15 @@ nvme_create_io_sq(struct nvme_ctrl *ctrl, struct nvme_sq *sq, u16 q_idx, struct if (!nvme_is_cqe_success(&cqe)) { dprintf(2, "create io sq failed: %08x %08x %08x %08x\n", cqe.dword[0], cqe.dword[1], cqe.dword[2], cqe.dword[3]); - return -1; + goto err_destroy_sq; } return 0; + +err_destroy_sq: + nvme_destroy_sq(sq); +err: + return -1; } /* Reads count sectors into buf. Returns DISK_RET_*. The buffer cannot cross @@ -384,17 +434,28 @@ nvme_io_readwrite(struct nvme_namespace *ns, u64 lba, char *buf, u16 count, return DISK_RET_SUCCESS; } - static int nvme_create_io_queues(struct nvme_ctrl *ctrl) { if (nvme_create_io_cq(ctrl, &ctrl->io_cq, 3)) - return -1; + goto err; if (nvme_create_io_sq(ctrl, &ctrl->io_sq, 2, &ctrl->io_cq)) - return -1; + goto err_free_cq; return 0; + + err_free_cq: + nvme_destroy_cq(&ctrl->io_cq); + err: + return -1; +} + +static void +nvme_destroy_io_queues(struct nvme_ctrl *ctrl) +{ + nvme_destroy_sq(&ctrl->io_sq); + nvme_destroy_cq(&ctrl->io_cq); } /* Waits for CSTS.RDY to match rdy. Returns 0 on success. */ @@ -426,6 +487,8 @@ nvme_wait_csts_rdy(struct nvme_ctrl *ctrl, unsigned rdy) static int nvme_controller_enable(struct nvme_ctrl *ctrl) { + int rc; + pci_enable_busmaster(ctrl->pci); /* Turn the controller off. */ @@ -437,18 +500,21 @@ nvme_controller_enable(struct nvme_ctrl *ctrl) ctrl->doorbell_stride = 4U << ((ctrl->reg->cap >> 32) & 0xF); - nvme_init_cq(ctrl, &ctrl->admin_cq, 1, - NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + rc = nvme_init_cq(ctrl, &ctrl->admin_cq, 1, + NVME_PAGE_SIZE / sizeof(struct nvme_cqe)); + if (rc) { + return -1; + } - nvme_init_sq(ctrl, &ctrl->admin_sq, 0, - NVME_PAGE_SIZE / sizeof(struct nvme_sqe), &ctrl->admin_cq); + rc = nvme_init_sq(ctrl, &ctrl->admin_sq, 0, + NVME_PAGE_SIZE / sizeof(struct nvme_sqe), &ctrl->admin_cq); + if (rc) { + goto err_destroy_admin_cq; + } ctrl->reg->aqa = ctrl->admin_cq.common.mask << 16 | ctrl->admin_sq.common.mask; - /* Create the admin queue pair */ - if (!ctrl->admin_sq.sqe || !ctrl->admin_cq.cqe) goto out_of_memory; - ctrl->reg->asq = (u32)ctrl->admin_sq.sqe; ctrl->reg->acq = (u32)ctrl->admin_cq.cqe; @@ -460,8 +526,9 @@ nvme_controller_enable(struct nvme_ctrl *ctrl) if (nvme_wait_csts_rdy(ctrl, 1)) { dprintf(2, "NVMe fatal error while enabling controller\n"); - goto failed; + goto err_destroy_admin_sq; } + /* The admin queue is set up and the controller is ready. Let's figure out what namespaces we have. */ @@ -469,10 +536,9 @@ nvme_controller_enable(struct nvme_ctrl *ctrl) if (!identify) { dprintf(2, "NVMe couldn't identify controller.\n"); - goto failed; + goto err_destroy_admin_sq; } - /* TODO Print model/serial info. */ dprintf(3, "NVMe has %u namespace%s.\n", identify->nn, (identify->nn == 1) ? "" : "s"); @@ -482,11 +548,14 @@ nvme_controller_enable(struct nvme_ctrl *ctrl) if ((ctrl->ns_count == 0) || nvme_create_io_queues(ctrl)) { /* No point to continue, if the controller says it doesn't have namespaces or we couldn't create I/O queues. */ - goto failed; + goto err_destroy_admin_sq; } ctrl->ns = malloc_fseg(sizeof(*ctrl->ns) * ctrl->ns_count); - if (!ctrl->ns) goto out_of_memory; + if (!ctrl->ns) { + warn_noalloc(); + goto err_destroy_ioq; + } memset(ctrl->ns, 0, sizeof(*ctrl->ns) * ctrl->ns_count); /* Populate namespace IDs */ @@ -498,12 +567,12 @@ nvme_controller_enable(struct nvme_ctrl *ctrl) dprintf(3, "NVMe initialization complete!\n"); return 0; - out_of_memory: - warn_noalloc(); - failed: - free(ctrl->admin_sq.sqe); - free(ctrl->admin_cq.cqe); - free(ctrl->ns); + err_destroy_ioq: + nvme_destroy_io_queues(ctrl); + err_destroy_admin_sq: + nvme_destroy_sq(&ctrl->admin_sq); + err_destroy_admin_cq: + nvme_destroy_cq(&ctrl->admin_cq); return -1; } @@ -524,13 +593,13 @@ nvme_controller_setup(void *opaque) if (~reg->cap & NVME_CAP_CSS_NVME) { dprintf(3, "Controller doesn't speak NVMe command set. Skipping.\n"); - return; + goto err; } struct nvme_ctrl *ctrl = malloc_high(sizeof(*ctrl)); if (!ctrl) { warn_noalloc(); - return; + goto err; } memset(ctrl, 0, sizeof(*ctrl)); @@ -539,9 +608,15 @@ nvme_controller_setup(void *opaque) ctrl->pci = pci; if (nvme_controller_enable(ctrl)) { - /* Initialization failed */ - free(ctrl); + goto err_free_ctrl; } + + return; + + err_free_ctrl: + free(ctrl); + err: + dprintf(2, "Failed to enable NVMe controller.\n"); } // Locate and init NVMe controllers -- 2.7.4

2 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

SeaBIOS October 2017