From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752415AbdHHJE5 (ORCPT ); Tue, 8 Aug 2017 05:04:57 -0400 Received: from mail-wr0-f178.google.com ([209.85.128.178]:35301 "EHLO mail-wr0-f178.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751962AbdHHJEv (ORCPT ); Tue, 8 Aug 2017 05:04:51 -0400 From: Suganath Prabu S To: JBottomley@Parallels.com, jejb@kernel.org, hch@infradead.org Cc: martin.petersen@oracle.com, linux-scsi@vger.kernel.org, Sathya.Prakash@broadcom.com, kashyap.desai@broadcom.com, linux-kernel@vger.kernel.org, suganath-prabu.subramani@broadcom.com, chaitra.basappa@broadcom.com, sreekanth.reddy@broadcom.com, linux-nvme@lists.infradead.org Subject: [PATCH v3 03/13] mpt3sas: SGL to PRP Translation for I/Os to NVMe devices Date: Tue, 8 Aug 2017 14:33:45 +0530 Message-Id: <1502183035-7441-4-git-send-email-suganath-prabu.subramani@broadcom.com> X-Mailer: git-send-email 2.0.2 In-Reply-To: <1502183035-7441-1-git-send-email-suganath-prabu.subramani@broadcom.com> References: <1502183035-7441-1-git-send-email-suganath-prabu.subramani@broadcom.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Added support for translating the SGLs associated with incoming commands either to IEE SGL or NVMe PRPs for NVMe devices. * The hardware translation of IEEE SGL to NVMe PRPs has limitation and if a command cannot be translated by hardware then it will go to firmware and the firmware needs to translate it. And this will have a performance reduction. To avoid that driver proactively checks whether the translation will be done in hardware or not, if not then driver try to translate inside the driver. Signed-off-by: Chaitra P B Signed-off-by: Suganath Prabu S --- drivers/scsi/mpt3sas/mpt3sas_base.c | 668 +++++++++++++++++++++++++++++- drivers/scsi/mpt3sas/mpt3sas_base.h | 44 ++- drivers/scsi/mpt3sas/mpt3sas_ctl.c | 1 + drivers/scsi/mpt3sas/mpt3sas_scsih.c | 14 +- drivers/scsi/mpt3sas/mpt3sas_warpdrive.c | 2 +- 5 files changed, 713 insertions(+), 16 deletions(-) diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c index 18039bb..d48f176 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_base.c +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c @@ -59,6 +59,7 @@ #include #include #include +#include /* To get host page size per arch */ #include @@ -1347,6 +1348,547 @@ _base_build_sg(struct MPT3SAS_ADAPTER *ioc, void *psge, /* IEEE format sgls */ /** + * _base_build_nvme_prp - This function is called for NVMe end devices to build + * a native SGL (NVMe PRP). The native SGL is built starting in the first PRP + * entry of the NVMe message (PRP1). If the data buffer is small enough to be + * described entirely using PRP1, then PRP2 is not used. If needed, PRP2 is + * used to describe a larger data buffer. If the data buffer is too large to + * describe using the two PRP entriess inside the NVMe message, then PRP1 + * describes the first data memory segment, and PRP2 contains a pointer to a PRP + * list located elsewhere in memory to describe the remaining data memory + * segments. The PRP list will be contiguous. + + * The native SGL for NVMe devices is a Physical Region Page (PRP). A PRP + * consists of a list of PRP entries to describe a number of noncontigous + * physical memory segments as a single memory buffer, just as a SGL does. Note + * however, that this function is only used by the IOCTL call, so the memory + * given will be guaranteed to be contiguous. There is no need to translate + * non-contiguous SGL into a PRP in this case. All PRPs will describe + * contiguous space that is one page size each. + * + * Each NVMe message contains two PRP entries. The first (PRP1) either contains + * a PRP list pointer or a PRP element, depending upon the command. PRP2 + * contains the second PRP element if the memory being described fits within 2 + * PRP entries, or a PRP list pointer if the PRP spans more than two entries. + * + * A PRP list pointer contains the address of a PRP list, structured as a linear + * array of PRP entries. Each PRP entry in this list describes a segment of + * physical memory. + * + * Each 64-bit PRP entry comprises an address and an offset field. The address + * always points at the beginning of a 4KB physical memory page, and the offset + * describes where within that 4KB page the memory segment begins. Only the + * first element in a PRP list may contain a non-zero offest, implying that all + * memory segments following the first begin at the start of a 4KB page. + * + * Each PRP element normally describes 4KB of physical memory, with exceptions + * for the first and last elements in the list. If the memory being described + * by the list begins at a non-zero offset within the first 4KB page, then the + * first PRP element will contain a non-zero offset indicating where the region + * begins within the 4KB page. The last memory segment may end before the end + * of the 4KB segment, depending upon the overall size of the memory being + * described by the PRP list. + * + * Since PRP entries lack any indication of size, the overall data buffer length + * is used to determine where the end of the data memory buffer is located, and + * how many PRP entries are required to describe it. + * + * @ioc: per adapter object + * @smid: system request message index for getting asscociated SGL + * @nvme_encap_request: the NVMe request msg frame pointer + * @data_out_dma: physical address for WRITES + * @data_out_sz: data xfer size for WRITES + * @data_in_dma: physical address for READS + * @data_in_sz: data xfer size for READS + * + * Returns nothing. + */ +static void +_base_build_nvme_prp(struct MPT3SAS_ADAPTER *ioc, u16 smid, + Mpi26NVMeEncapsulatedRequest_t *nvme_encap_request, + dma_addr_t data_out_dma, size_t data_out_sz, dma_addr_t data_in_dma, + size_t data_in_sz) +{ + int prp_size = NVME_PRP_SIZE; + u64 *prp_entry, *prp1_entry, *prp2_entry, *prp_entry_phys; + u64 *prp_page, *prp_page_phys; + u32 offset, entry_len; + u32 page_mask_result, page_mask; + dma_addr_t paddr; + size_t length; + + /* + * Not all commands require a data transfer. If no data, just return + * without constructing any PRP. + */ + if (!data_in_sz && !data_out_sz) + return; + /* + * Set pointers to PRP1 and PRP2, which are in the NVMe command. + * PRP1 is located at a 24 byte offset from the start of the NVMe + * command. Then set the current PRP entry pointer to PRP1. + */ + prp1_entry = (u64 *)(nvme_encap_request->NVMe_Command + + NVME_CMD_PRP1_OFFSET); + prp2_entry = (u64 *)(nvme_encap_request->NVMe_Command + + NVME_CMD_PRP2_OFFSET); + prp_entry = prp1_entry; + /* + * For the PRP entries, use the specially allocated buffer of + * contiguous memory. + */ + prp_page = (u64 *)mpt3sas_base_get_pcie_sgl(ioc, smid); + prp_page_phys = (u64 *)mpt3sas_base_get_pcie_sgl_dma(ioc, smid); + + /* + * Check if we are within 1 entry of a page boundary we don't + * want our first entry to be a PRP List entry. + */ + page_mask = ioc->page_size - 1; + page_mask_result = (uintptr_t)((u8 *)prp_page + prp_size) & page_mask; + if (!page_mask_result) { + /* Bump up to next page boundary. */ + prp_page = (u64 *)((u8 *)prp_page + prp_size); + prp_page_phys = (u64 *)((u8 *)prp_page_phys + prp_size); + } + + /* + * Set PRP physical pointer, which initially points to the current PRP + * DMA memory page. + */ + prp_entry_phys = prp_page_phys; + + /* Get physical address and length of the data buffer. */ + if (data_in_sz) { + paddr = data_in_dma; + length = data_in_sz; + } else { + paddr = data_out_dma; + length = data_out_sz; + } + + /* Loop while the length is not zero. */ + while (length) { + /* + * Check if we need to put a list pointer here if we are at + * page boundary - prp_size (8 bytes). + */ + page_mask_result = + (uintptr_t)((u8 *)prp_entry_phys + prp_size) & page_mask; + if (!page_mask_result) { + /* + * This is the last entry in a PRP List, so we need to + * put a PRP list pointer here. What this does is: + * - bump the current memory pointer to the next + * address, which will be the next full page. + * - set the PRP Entry to point to that page. This + * is now the PRP List pointer. + * - bump the PRP Entry pointer the start of the + * next page. Since all of this PRP memory is + * contiguous, no need to get a new page - it's + * just the next address. + */ + prp_entry_phys++; + *prp_entry = cpu_to_le64((uintptr_t)prp_entry_phys); + prp_entry++; + } + + /* Need to handle if entry will be part of a page. */ + offset = (u32)paddr & page_mask; + entry_len = ioc->page_size - offset; + + if (prp_entry == prp1_entry) { + /* + * Must fill in the first PRP pointer (PRP1) before + * moving on. + */ + *prp1_entry = cpu_to_le64((u64)paddr); + + /* + * Now point to the second PRP entry within the + * command (PRP2). + */ + prp_entry = prp2_entry; + } else if (prp_entry == prp2_entry) { + /* + * Should the PRP2 entry be a PRP List pointer or just + * a regular PRP pointer? If there is more than one + * more page of data, must use a PRP List pointer. + */ + if (length > ioc->page_size) { + /* + * PRP2 will contain a PRP List pointer because + * more PRP's are needed with this command. The + * list will start at the beginning of the + * contiguous buffer. + */ + *prp2_entry = + cpu_to_le64((uintptr_t)prp_entry_phys); + + /* + * The next PRP Entry will be the start of the + * first PRP List. + */ + prp_entry = prp_page; + } else { + /* + * After this, the PRP Entries are complete. + * This command uses 2 PRP's and no PRP list. + */ + *prp2_entry = cpu_to_le64((u64)paddr); + } + } else { + /* + * Put entry in list and bump the addresses. + * + * After PRP1 and PRP2 are filled in, this will fill in + * all remaining PRP entries in a PRP List, one per + * each time through the loop. + */ + *prp_entry = cpu_to_le64((u64)paddr); + prp_entry++; + prp_entry_phys++; + } + + /* + * Bump the phys address of the command's data buffer by the + * entry_len. + */ + paddr += entry_len; + + /* Decrement length accounting for last partial page. */ + if (entry_len > length) + length = 0; + else + length -= entry_len; + } +} + +u32 base_mod64(u64 dividend, u32 divisor) +{ + u64 d; + u32 remainder; + + if (!divisor) + pr_err(KERN_ERR "mpt3sas : DIVISOR is zero, in div fn\n"); + d = dividend; + remainder = do_div(d, divisor); + return remainder; +} + +/** + * base_make_prp_nvme - + * Prepare PRPs(Physical Region Page)- SGLs specific to NVMe drives only + * + * @ioc: per adapter object + * @scmd: SCSI command from the mid-layer + * @mpi_request: mpi request + * @smid: msg Index + * @sge_count: scatter gather element count. + * + * Returns: true: PRPs are built + * false: IEEE SGLs needs to be built + */ +void +base_make_prp_nvme(struct MPT3SAS_ADAPTER *ioc, + struct scsi_cmnd *scmd, + Mpi25SCSIIORequest_t *mpi_request, + u16 smid, int sge_count) +{ + int sge_len, offset, num_prp_in_chain = 0; + Mpi25IeeeSgeChain64_t *main_chain_element, *ptr_first_sgl; + u64 *curr_buff; + dma_addr_t msg_phys; + u64 sge_addr; + u32 page_mask, page_mask_result; + struct scatterlist *sg_scmd; + u32 first_prp_len; + int data_len = scsi_bufflen(scmd); + u32 nvme_pg_size; + + nvme_pg_size = max_t(u32, ioc->page_size, NVME_PRP_PAGE_SIZE); + /* + * Nvme has a very convoluted prp format. One prp is required + * for each page or partial page. Driver need to split up OS sg_list + * entries if it is longer than one page or cross a page + * boundary. Driver also have to insert a PRP list pointer entry as + * the last entry in each physical page of the PRP list. + * + * NOTE: The first PRP "entry" is actually placed in the first + * SGL entry in the main message as IEEE 64 format. The 2nd + * entry in the main message is the chain element, and the rest + * of the PRP entries are built in the contiguous pcie buffer. + */ + page_mask = nvme_pg_size - 1; + + /* + * Native SGL is needed. + * Put a chain element in main message frame that points to the first + * chain buffer. + * + * NOTE: The ChainOffset field must be 0 when using a chain pointer to + * a native SGL. + */ + + /* Set main message chain element pointer */ + main_chain_element = (pMpi25IeeeSgeChain64_t)&mpi_request->SGL; + /* + * For NVMe the chain element needs to be the 2nd SG entry in the main + * message. + */ + main_chain_element = (Mpi25IeeeSgeChain64_t *) + ((u8 *)main_chain_element + sizeof(MPI25_IEEE_SGE_CHAIN64)); + + /* + * For the PRP entries, use the specially allocated buffer of + * contiguous memory. Normal chain buffers can't be used + * because each chain buffer would need to be the size of an OS + * page (4k). + */ + curr_buff = mpt3sas_base_get_pcie_sgl(ioc, smid); + msg_phys = (dma_addr_t)mpt3sas_base_get_pcie_sgl_dma(ioc, smid); + + main_chain_element->Address = cpu_to_le64(msg_phys); + main_chain_element->NextChainOffset = 0; + main_chain_element->Flags = MPI2_IEEE_SGE_FLAGS_CHAIN_ELEMENT | + MPI2_IEEE_SGE_FLAGS_SYSTEM_ADDR | + MPI26_IEEE_SGE_FLAGS_NSF_NVME_PRP; + + /* Build first prp, sge need not to be page aligned*/ + ptr_first_sgl = (pMpi25IeeeSgeChain64_t)&mpi_request->SGL; + sg_scmd = scsi_sglist(scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + + offset = (u32)(sge_addr & page_mask); + first_prp_len = nvme_pg_size - offset; + + ptr_first_sgl->Address = cpu_to_le64(sge_addr); + ptr_first_sgl->Length = cpu_to_le32(first_prp_len); + + data_len -= first_prp_len; + + if (sge_len > first_prp_len) { + sge_addr += first_prp_len; + sge_len -= first_prp_len; + } else if (sge_len == first_prp_len) { + sg_scmd = sg_next(sg_scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + } + + for (;;) { + offset = (u32)(sge_addr & page_mask); + + /* Put PRP pointer due to page boundary*/ + page_mask_result = (uintptr_t)(curr_buff + 1) & page_mask; + if (unlikely(!page_mask_result)) { + scmd_printk(KERN_NOTICE, + scmd, "page boundary curr_buff: 0x%p\n", + curr_buff); + msg_phys += 8; + *curr_buff = cpu_to_le64(msg_phys); + curr_buff++; + num_prp_in_chain++; + } + + *curr_buff = cpu_to_le64(sge_addr); + curr_buff++; + msg_phys += 8; + num_prp_in_chain++; + + sge_addr += nvme_pg_size; + sge_len -= nvme_pg_size; + data_len -= nvme_pg_size; + + if (data_len <= 0) + break; + + if (sge_len > 0) + continue; + + sg_scmd = sg_next(sg_scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + } + + main_chain_element->Length = + cpu_to_le32(num_prp_in_chain * sizeof(u64)); + return; +} + +static bool +base_is_prp_possible(struct MPT3SAS_ADAPTER *ioc, + struct _pcie_device *pcie_device, struct scsi_cmnd *scmd, int sge_count) +{ + u32 i; + u32 data_length = 0; + struct scatterlist *sg_scmd; + bool build_prp = false; + u32 nvme_pg_size, page_mask; + u32 first_page_offset, first_page_data_size, end_residual; + u64 *msg_phys; + + nvme_pg_size = max_t(u32, ioc->page_size, + NVME_PRP_PAGE_SIZE); + data_length = cpu_to_le32(scsi_bufflen(scmd)); + sg_scmd = scsi_sglist(scmd); + + /* Create page_mask (to get offset within page) */ + page_mask = ioc->page_size - 1; + + /* + * For NVMe, check if the data transfer length exceeds the MDTS (Maximum + * Data Transfer Size) for this device. If so, just return 1 so a + * normal IEEE SGL is built. F/W will break up the I/O into multiple + * I/O's. [nvme_mdts = 0 means unlimited] + */ + + if ((pcie_device->nvme_mdts > 0) && + (data_length > pcie_device->nvme_mdts)) + return build_prp; + + /* + ** NVMe uses one PRP for each page (or part of a page) + ** look at the data length - if 4 pages or less then IEEE is OK + ** if > 5 pages then we need to build a native SGL + ** if > 4 and <= 5 pages, then check physical address of 1st SG entry + ** if this first size in the page is >= the residual beyond 4 pages + ** then use IEEE, otherwise use native SGL + **/ + + if (data_length > (nvme_pg_size * 5)) { + build_prp = true; + } else if ((data_length > (nvme_pg_size * 4)) && + (data_length <= (nvme_pg_size * 5))) { + msg_phys = (u64 *) sg_dma_address(sg_scmd); + first_page_offset = ((u32)(u64)msg_phys & page_mask); + first_page_data_size = ioc->page_size - + first_page_offset; + end_residual = data_length % ioc->page_size; + /* + ** If offset into first page pushes the end of + ** the data beyond end of the 5th page, we need + ** the extra PRP list + **/ + if (first_page_data_size < end_residual) + build_prp = true; + + /* check if 1st SG entry size is < residual beyond 4 pages */ + if (sg_dma_len(sg_scmd) < (data_length - (nvme_pg_size * 4))) + build_prp = true; + } + + /* + ** Below code detects gaps/holes in IO data buffers. + ** What does holes/gaps mean? + ** Any SGE except first one in a SGL starts at non NVME page size + ** aligned address OR Any SGE except last one in a SGL ends at + ** non NVME page size boundary. + ** + ** Driver has already informed block layer by setting boundary rules + ** for bio merging done at NVME page size boundary calling kernel API + ** blk_queue_virt_boundary inside slave_config. + ** Still there is possibility of IO coming with holes to driver because + ** of IO merging done by IO scheduler. + ** + ** With SCSI BLK MQ enabled, there will be no IO with holes as there + ** is no IO scheduling so no IO merging. + ** + ** With SCSI BLK MQ disabled, IO scheduler may attempt to merge IOs and + ** then sending IOs with holes. + ** + ** Though driver can request block layer to disable IO merging by + ** calling queue_flag_set_unlocked(QUEUE_FLAG_NOMERGES, + ** sdev->request_queue) but user may tune sysfs parameter- nomerges + ** again to 0 or 1. + ** + ** If in future IO scheduling is enabled with SCSI BLK MQ, + ** this algorithm to detect holes will be required in driver + ** for SCSI BLK MQ enabled case as well. + ** + **/ + scsi_for_each_sg(scmd, sg_scmd, sge_count, i) { + if ((i != 0) && (i != (sge_count - 1))) { + if (base_mod64(sg_dma_len(sg_scmd), nvme_pg_size) || + base_mod64(sg_dma_address(sg_scmd), + nvme_pg_size)) { + build_prp = false; + break; + } + } + + if ((sge_count > 1) && (i == 0)) { + if ((base_mod64((sg_dma_address(sg_scmd) + + sg_dma_len(sg_scmd)), nvme_pg_size))) { + build_prp = false; + break; + } + } + + if ((sge_count > 1) && (i == (sge_count - 1))) { + if (base_mod64(sg_dma_address(sg_scmd), nvme_pg_size)) { + build_prp = false; + break; + } + } + } + return build_prp; +} + +/** + * _base_check_pcie_native_sgl - This function is called for PCIe end devices to + * determine if the driver needs to build a native SGL. If so, that native + * SGL is built in the special contiguous buffers allocated especially for + * PCIe SGL creation. If the driver will not build a native SGL, return + * TRUE and a normal IEEE SGL will be built. Currently this routine + * supports NVMe. + * @ioc: per adapter object + * @mpi_request: mf request pointer + * @smid: system request message index + * @scmd: scsi command + * @pcie_device: points to the PCIe device's info + * + * Returns 0 if native SGL was built, 1 if no SGL was built + */ +static int +_base_check_pcie_native_sgl(struct MPT3SAS_ADAPTER *ioc, + Mpi25SCSIIORequest_t *mpi_request, u16 smid, struct scsi_cmnd *scmd, + struct _pcie_device *pcie_device) +{ + struct scatterlist *sg_scmd; + int sges_left; + + /* Get the SG list pointer and info. */ + sg_scmd = scsi_sglist(scmd); + sges_left = scsi_dma_map(scmd); + if (sges_left < 0) { + sdev_printk(KERN_ERR, scmd->device, + "scsi_dma_map failed: request for %d bytes!\n", + scsi_bufflen(scmd)); + return 1; + } + + /* Check if we need to build a native SG list. */ + if (base_is_prp_possible(ioc, pcie_device, + scmd, sges_left) == 0) { + /* We built a native SG list, just return. */ + goto out; + } + + /* + * Build native NVMe PRP. + */ + base_make_prp_nvme(ioc, scmd, mpi_request, + smid, sges_left); + + return 0; +out: + scsi_dma_unmap(scmd); + return 1; +} + +/** * _base_add_sg_single_ieee - add sg element for IEEE format * @paddr: virtual address for SGE * @flags: SGE flags @@ -1391,9 +1933,11 @@ _base_build_zero_len_sge_ieee(struct MPT3SAS_ADAPTER *ioc, void *paddr) /** * _base_build_sg_scmd - main sg creation routine + * pcie_device is unused here! * @ioc: per adapter object * @scmd: scsi command * @smid: system request message index + * @unused: unused pcie_device pointer * Context: none. * * The main routine that builds scatter gather table from a given @@ -1403,7 +1947,7 @@ _base_build_zero_len_sge_ieee(struct MPT3SAS_ADAPTER *ioc, void *paddr) */ static int _base_build_sg_scmd(struct MPT3SAS_ADAPTER *ioc, - struct scsi_cmnd *scmd, u16 smid) + struct scsi_cmnd *scmd, u16 smid, struct _pcie_device *unused) { Mpi2SCSIIORequest_t *mpi_request; dma_addr_t chain_dma; @@ -1537,6 +2081,8 @@ _base_build_sg_scmd(struct MPT3SAS_ADAPTER *ioc, * @ioc: per adapter object * @scmd: scsi command * @smid: system request message index + * @pcie_device: Pointer to pcie_device. If set, the pcie native sgl will be + * constructed on need. * Context: none. * * The main routine that builds scatter gather table from a given @@ -1546,9 +2092,9 @@ _base_build_sg_scmd(struct MPT3SAS_ADAPTER *ioc, */ static int _base_build_sg_scmd_ieee(struct MPT3SAS_ADAPTER *ioc, - struct scsi_cmnd *scmd, u16 smid) + struct scsi_cmnd *scmd, u16 smid, struct _pcie_device *pcie_device) { - Mpi2SCSIIORequest_t *mpi_request; + Mpi25SCSIIORequest_t *mpi_request; dma_addr_t chain_dma; struct scatterlist *sg_scmd; void *sg_local, *chain; @@ -1571,6 +2117,13 @@ _base_build_sg_scmd_ieee(struct MPT3SAS_ADAPTER *ioc, chain_sgl_flags = MPI2_IEEE_SGE_FLAGS_CHAIN_ELEMENT | MPI2_IEEE_SGE_FLAGS_SYSTEM_ADDR; + /* Check if we need to build a native SG list. */ + if ((pcie_device) && (_base_check_pcie_native_sgl(ioc, mpi_request, + smid, scmd, pcie_device) == 0)) { + /* We built a native SG list, just return. */ + return 0; + } + sg_scmd = scsi_sglist(scmd); sges_left = scsi_dma_map(scmd); if (sges_left < 0) { @@ -1582,12 +2135,12 @@ _base_build_sg_scmd_ieee(struct MPT3SAS_ADAPTER *ioc, sg_local = &mpi_request->SGL; sges_in_segment = (ioc->request_sz - - offsetof(Mpi2SCSIIORequest_t, SGL))/ioc->sge_size_ieee; + offsetof(Mpi25SCSIIORequest_t, SGL))/ioc->sge_size_ieee; if (sges_left <= sges_in_segment) goto fill_in_last_segment; mpi_request->ChainOffset = (sges_in_segment - 1 /* chain element */) + - (offsetof(Mpi2SCSIIORequest_t, SGL)/ioc->sge_size_ieee); + (offsetof(Mpi25SCSIIORequest_t, SGL)/ioc->sge_size_ieee); /* fill in main message segment when there is a chain following */ while (sges_in_segment > 1) { @@ -2267,6 +2820,33 @@ mpt3sas_base_get_sense_buffer_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid) } /** + * mpt3sas_base_get_pcie_sgl - obtain a PCIe SGL virt addr + * @ioc: per adapter object + * @smid: system request message index + * + * Returns virt pointer to a PCIe SGL. + */ +void * +mpt3sas_base_get_pcie_sgl(struct MPT3SAS_ADAPTER *ioc, u16 smid) +{ + return (void *)(ioc->scsi_lookup[smid - 1].pcie_sg_list.pcie_sgl); +} + +/** + * mpt3sas_base_get_pcie_sgl_dma - obtain a PCIe SGL dma addr + * @ioc: per adapter object + * @smid: system request message index + * + * Returns phys pointer to the address of the PCIe buffer. + */ +void * +mpt3sas_base_get_pcie_sgl_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid) +{ + return (void *)(uintptr_t) + (ioc->scsi_lookup[smid - 1].pcie_sg_list.pcie_sgl_dma); +} + +/** * mpt3sas_base_get_reply_virt_addr - obtain reply frames virt address * @ioc: per adapter object * @phys_addr: lower 32 physical addr of the reply @@ -2945,6 +3525,11 @@ _base_display_ioc_capabilities(struct MPT3SAS_ADAPTER *ioc) _base_display_OEMs_branding(ioc); + if (ioc->facts.ProtocolFlags & MPI2_IOCFACTS_PROTOCOL_NVME_DEVICES) { + pr_info("%sNVMe", i ? "," : ""); + i++; + } + pr_info(MPT3SAS_FMT "Protocol=(", ioc->name); if (ioc->facts.ProtocolFlags & MPI2_IOCFACTS_PROTOCOL_SCSI_INITIATOR) { @@ -3249,6 +3834,17 @@ _base_release_memory_pools(struct MPT3SAS_ADAPTER *ioc) kfree(ioc->reply_post); } + if (ioc->pcie_sgl_dma_pool) { + for (i = 0; i < ioc->scsiio_depth; i++) { + if (ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl) + pci_pool_free(ioc->pcie_sgl_dma_pool, + ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl, + ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl_dma); + } + if (ioc->pcie_sgl_dma_pool) + pci_pool_destroy(ioc->pcie_sgl_dma_pool); + } + if (ioc->config_page) { dexitprintk(ioc, pr_info(MPT3SAS_FMT "config_page(0x%p): free\n", ioc->name, @@ -3291,7 +3887,7 @@ _base_allocate_memory_pools(struct MPT3SAS_ADAPTER *ioc) u16 chains_needed_per_io; u32 sz, total_sz, reply_post_free_sz; u32 retry_sz; - u16 max_request_credit; + u16 max_request_credit, nvme_blocks_needed; unsigned short sg_tablesize; u16 sge_size; int i; @@ -3627,7 +4223,52 @@ _base_allocate_memory_pools(struct MPT3SAS_ADAPTER *ioc) "internal(0x%p): depth(%d), start smid(%d)\n", ioc->name, ioc->internal, ioc->internal_depth, ioc->internal_smid)); + /* + * The number of NVMe page sized blocks needed is: + * (((sg_tablesize * 8) - 1) / (page_size - 8)) + 1 + * ((sg_tablesize * 8) - 1) is the max PRP's minus the first PRP entry + * that is placed in the main message frame. 8 is the size of each PRP + * entry or PRP list pointer entry. 8 is subtracted from page_size + * because of the PRP list pointer entry at the end of a page, so this + * is not counted as a PRP entry. The 1 added page is a round up. + * + * To avoid allocation failures due to the amount of memory that could + * be required for NVMe PRP's, only each set of NVMe blocks will be + * contiguous, so a new set is allocated for each possible I/O. + */ + if (ioc->facts.ProtocolFlags & MPI2_IOCFACTS_PROTOCOL_NVME_DEVICES) { + nvme_blocks_needed = + (ioc->shost->sg_tablesize * NVME_PRP_SIZE) - 1; + nvme_blocks_needed /= (ioc->page_size - NVME_PRP_SIZE); + nvme_blocks_needed++; + + sz = nvme_blocks_needed * ioc->page_size; + ioc->pcie_sgl_dma_pool = + pci_pool_create("PCIe SGL pool", ioc->pdev, sz, 16, 0); + if (!ioc->pcie_sgl_dma_pool) { + pr_info(MPT3SAS_FMT + "PCIe SGL pool: pci_pool_create failed\n", + ioc->name); + goto out; + } + for (i = 0; i < ioc->scsiio_depth; i++) { + ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl = + pci_pool_alloc(ioc->pcie_sgl_dma_pool, + GFP_KERNEL, + &ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl_dma); + if (!ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl) { + pr_info(MPT3SAS_FMT + "PCIe SGL pool: pci_pool_alloc failed\n", + ioc->name); + goto out; + } + } + dinitprintk(ioc, pr_info(MPT3SAS_FMT "PCIe sgl pool depth(%d), " + "element_size(%d), pool_size(%d kB)\n", ioc->name, + ioc->scsiio_depth, sz, (sz * ioc->scsiio_depth)/1024)); + total_sz += sz * ioc->scsiio_depth; + } /* sense buffers, 4 byte align */ sz = ioc->scsiio_depth * SCSI_SENSE_BUFFERSIZE; ioc->sense_dma_pool = pci_pool_create("sense pool", ioc->pdev, sz, 4, @@ -4472,6 +5113,19 @@ _base_get_ioc_facts(struct MPT3SAS_ADAPTER *ioc) le16_to_cpu(mpi_reply.HighPriorityCredit); facts->ReplyFrameSize = mpi_reply.ReplyFrameSize; facts->MaxDevHandle = le16_to_cpu(mpi_reply.MaxDevHandle); + facts->CurrentHostPageSize = mpi_reply.CurrentHostPageSize; + + /* + * Get the Page Size from IOC Facts. If it's 0, default to 4k. + */ + ioc->page_size = 1 << facts->CurrentHostPageSize; + if (ioc->page_size == 1) { + pr_info(MPT3SAS_FMT "CurrentHostPageSize is 0: Setting " + "default host page size to 4k\n", ioc->name); + ioc->page_size = 1 << MPT3SAS_HOST_PAGE_SIZE_4K; + } + dinitprintk(ioc, pr_info(MPT3SAS_FMT "CurrentHostPageSize(%d)\n", + ioc->name, facts->CurrentHostPageSize)); dinitprintk(ioc, pr_info(MPT3SAS_FMT "hba queue depth(%d), max chains per io(%d)\n", @@ -4511,6 +5165,7 @@ _base_send_ioc_init(struct MPT3SAS_ADAPTER *ioc) mpi_request.VP_ID = 0; mpi_request.MsgVersion = cpu_to_le16(ioc->hba_mpi_version_belonged); mpi_request.HeaderVersion = cpu_to_le16(MPI2_HEADER_VERSION); + mpi_request.HostPageSize = MPT3SAS_HOST_PAGE_SIZE_4K; if (_base_is_controller_msix_enabled(ioc)) mpi_request.HostMSIxVectors = ioc->reply_queue_count; @@ -5379,6 +6034,7 @@ mpt3sas_base_attach(struct MPT3SAS_ADAPTER *ioc) */ ioc->build_sg_scmd = &_base_build_sg_scmd_ieee; ioc->build_sg = &_base_build_sg_ieee; + ioc->build_nvme_prp = &_base_build_nvme_prp; ioc->build_zero_len_sge = &_base_build_zero_len_sge_ieee; ioc->sge_size_ieee = sizeof(Mpi2IeeeSgeSimple64_t); diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.h b/drivers/scsi/mpt3sas/mpt3sas_base.h index c522057..4caa91e 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_base.h +++ b/drivers/scsi/mpt3sas/mpt3sas_base.h @@ -54,6 +54,7 @@ #include "mpi/mpi2_raid.h" #include "mpi/mpi2_tool.h" #include "mpi/mpi2_sas.h" +#include "mpi/mpi2_pci.h" #include #include @@ -113,6 +114,7 @@ #define MPT3SAS_RAID_QUEUE_DEPTH 128 #define MPT3SAS_RAID_MAX_SECTORS 8192 +#define MPT3SAS_HOST_PAGE_SIZE_4K 12 #define MPT_NAME_LENGTH 32 /* generic length of strings */ #define MPT_STRING_LENGTH 64 @@ -131,6 +133,15 @@ #define DEFAULT_NUM_FWCHAIN_ELEMTS 8 /* + * NVMe defines + */ +#define NVME_PRP_SIZE 8 /* PRP size */ +#define NVME_CMD_PRP1_OFFSET 24 /* PRP1 offset in NVMe cmd */ +#define NVME_CMD_PRP2_OFFSET 32 /* PRP2 offset in NVMe cmd */ +#define NVME_ERROR_RESPONSE_SIZE 16 /* Max NVME Error Response */ +#define NVME_PRP_PAGE_SIZE 4096 /* Page size */ + +/* * reset phases */ #define MPT3_IOC_PRE_RESET 1 /* prior to host reset */ @@ -731,6 +742,16 @@ enum reset_type { }; /** + * struct pcie_sg_list - PCIe SGL buffer (contiguous per I/O) + * @pcie_sgl: PCIe native SGL for NVMe devices + * @pcie_sgl_dma: physical address + */ +struct pcie_sg_list { + void *pcie_sgl; + dma_addr_t pcie_sgl_dma; +}; + +/** * struct chain_tracker - firmware chain tracker * @chain_buffer: chain buffer * @chain_buffer_dma: physical address @@ -756,6 +777,7 @@ struct scsiio_tracker { struct scsi_cmnd *scmd; u8 cb_idx; u8 direct_io; + struct pcie_sg_list pcie_sg_list; struct list_head chain_list; struct list_head tracker_list; u16 msix_io; @@ -829,13 +851,19 @@ typedef void (*MPT_ADD_SGE)(void *paddr, u32 flags_length, dma_addr_t dma_addr); /* SAS3.0 support */ typedef int (*MPT_BUILD_SG_SCMD)(struct MPT3SAS_ADAPTER *ioc, - struct scsi_cmnd *scmd, u16 smid); + struct scsi_cmnd *scmd, u16 smid, struct _pcie_device *pcie_device); typedef void (*MPT_BUILD_SG)(struct MPT3SAS_ADAPTER *ioc, void *psge, dma_addr_t data_out_dma, size_t data_out_sz, dma_addr_t data_in_dma, size_t data_in_sz); typedef void (*MPT_BUILD_ZERO_LEN_SGE)(struct MPT3SAS_ADAPTER *ioc, void *paddr); +/* SAS3.5 support */ +typedef void (*NVME_BUILD_PRP)(struct MPT3SAS_ADAPTER *ioc, u16 smid, + Mpi26NVMeEncapsulatedRequest_t *nvme_encap_request, + dma_addr_t data_out_dma, size_t data_out_sz, dma_addr_t data_in_dma, + size_t data_in_sz); + /* To support atomic and non atomic descriptors*/ typedef void (*PUT_SMID_IO_FP_HIP) (struct MPT3SAS_ADAPTER *ioc, u16 smid, u16 funcdep); @@ -878,6 +906,7 @@ struct mpt3sas_facts { u16 MaxDevHandle; u16 MaxPersistentEntries; u16 MinDevHandle; + u8 CurrentHostPageSize; }; struct mpt3sas_port_facts { @@ -1149,6 +1178,9 @@ struct MPT3SAS_ADAPTER { MPT_BUILD_SG build_sg_mpi; MPT_BUILD_ZERO_LEN_SGE build_zero_len_sge_mpi; + /* function ptr for NVMe PRP elements only */ + NVME_BUILD_PRP build_nvme_prp; + /* event log */ u32 event_type[MPI2_EVENT_NOTIFY_EVENTMASK_WORDS]; u32 event_context; @@ -1217,6 +1249,11 @@ struct MPT3SAS_ADAPTER { int pending_io_count; wait_queue_head_t reset_wq; + /* PCIe SGL */ + struct dma_pool *pcie_sgl_dma_pool; + /* Host Page Size */ + u32 page_size; + /* chain */ struct chain_tracker *chain_lookup; struct list_head free_chain_list; @@ -1350,7 +1387,8 @@ void *mpt3sas_base_get_msg_frame(struct MPT3SAS_ADAPTER *ioc, u16 smid); void *mpt3sas_base_get_sense_buffer(struct MPT3SAS_ADAPTER *ioc, u16 smid); __le32 mpt3sas_base_get_sense_buffer_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid); - +void *mpt3sas_base_get_pcie_sgl(struct MPT3SAS_ADAPTER *ioc, u16 smid); +void *mpt3sas_base_get_pcie_sgl_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid); void mpt3sas_base_sync_reply_irqs(struct MPT3SAS_ADAPTER *ioc); /* hi-priority queue */ @@ -1564,7 +1602,7 @@ void mpt3sas_scsi_direct_io_set(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 direct_io); void mpt3sas_setup_direct_io(struct MPT3SAS_ADAPTER *ioc, struct scsi_cmnd *scmd, - struct _raid_device *raid_device, Mpi2SCSIIORequest_t *mpi_request, + struct _raid_device *raid_device, Mpi25SCSIIORequest_t *mpi_request, u16 smid); /* NCQ Prio Handling Check */ diff --git a/drivers/scsi/mpt3sas/mpt3sas_ctl.c b/drivers/scsi/mpt3sas/mpt3sas_ctl.c index bdffb69..0c18831 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_ctl.c +++ b/drivers/scsi/mpt3sas/mpt3sas_ctl.c @@ -299,6 +299,7 @@ mpt3sas_ctl_done(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 msix_index, } } } + _ctl_display_some_debug(ioc, smid, "ctl_done", mpi_reply); ioc->ctl_cmds.status &= ~MPT3_CMD_PENDING; complete(&ioc->ctl_cmds.done); diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c index 4b6b0fa..3e6dd4b 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c +++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c @@ -4244,7 +4244,7 @@ _scsih_flush_running_cmds(struct MPT3SAS_ADAPTER *ioc) */ static void _scsih_setup_eedp(struct MPT3SAS_ADAPTER *ioc, struct scsi_cmnd *scmd, - Mpi2SCSIIORequest_t *mpi_request) + Mpi25SCSIIORequest_t *mpi_request) { u16 eedp_flags; unsigned char prot_op = scsi_get_prot_op(scmd); @@ -4347,7 +4347,8 @@ scsih_qcmd(struct Scsi_Host *shost, struct scsi_cmnd *scmd) struct _raid_device *raid_device; struct request *rq = scmd->request; int class; - Mpi2SCSIIORequest_t *mpi_request; + Mpi25SCSIIORequest_t *mpi_request; + struct _pcie_device *pcie_device = NULL; u32 mpi_control; u16 smid; u16 handle; @@ -4435,7 +4436,7 @@ scsih_qcmd(struct Scsi_Host *shost, struct scsi_cmnd *scmd) goto out; } mpi_request = mpt3sas_base_get_msg_frame(ioc, smid); - memset(mpi_request, 0, sizeof(Mpi2SCSIIORequest_t)); + memset(mpi_request, 0, ioc->request_sz); _scsih_setup_eedp(ioc, scmd, mpi_request); if (scmd->cmd_len == 32) @@ -4454,13 +4455,14 @@ scsih_qcmd(struct Scsi_Host *shost, struct scsi_cmnd *scmd) mpi_request->SenseBufferLength = SCSI_SENSE_BUFFERSIZE; mpi_request->SenseBufferLowAddress = mpt3sas_base_get_sense_buffer_dma(ioc, smid); - mpi_request->SGLOffset0 = offsetof(Mpi2SCSIIORequest_t, SGL) / 4; + mpi_request->SGLOffset0 = offsetof(Mpi25SCSIIORequest_t, SGL) / 4; int_to_scsilun(sas_device_priv_data->lun, (struct scsi_lun *) mpi_request->LUN); memcpy(mpi_request->CDB.CDB32, scmd->cmnd, scmd->cmd_len); if (mpi_request->DataLength) { - if (ioc->build_sg_scmd(ioc, scmd, smid)) { + pcie_device = sas_target_priv_data->pcie_dev; + if (ioc->build_sg_scmd(ioc, scmd, smid, pcie_device)) { mpt3sas_base_free_smid(ioc, smid); goto out; } @@ -4931,7 +4933,7 @@ out_unlock: static u8 _scsih_io_done(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 msix_index, u32 reply) { - Mpi2SCSIIORequest_t *mpi_request; + Mpi25SCSIIORequest_t *mpi_request; Mpi2SCSIIOReply_t *mpi_reply; struct scsi_cmnd *scmd; u16 ioc_status; diff --git a/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c b/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c index 540bd50..ced7d9f 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c +++ b/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c @@ -299,7 +299,7 @@ mpt3sas_scsi_direct_io_set(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 direct_io) */ void mpt3sas_setup_direct_io(struct MPT3SAS_ADAPTER *ioc, struct scsi_cmnd *scmd, - struct _raid_device *raid_device, Mpi2SCSIIORequest_t *mpi_request, + struct _raid_device *raid_device, Mpi25SCSIIORequest_t *mpi_request, u16 smid) { sector_t v_lba, p_lba, stripe_off, column, io_size; -- 1.7.1 From mboxrd@z Thu Jan 1 00:00:00 1970 From: suganath-prabu.subramani@broadcom.com (Suganath Prabu S) Date: Tue, 8 Aug 2017 14:33:45 +0530 Subject: [PATCH v3 03/13] mpt3sas: SGL to PRP Translation for I/Os to NVMe devices In-Reply-To: <1502183035-7441-1-git-send-email-suganath-prabu.subramani@broadcom.com> References: <1502183035-7441-1-git-send-email-suganath-prabu.subramani@broadcom.com> Message-ID: <1502183035-7441-4-git-send-email-suganath-prabu.subramani@broadcom.com> * Added support for translating the SGLs associated with incoming commands either to IEE SGL or NVMe PRPs for NVMe devices. * The hardware translation of IEEE SGL to NVMe PRPs has limitation and if a command cannot be translated by hardware then it will go to firmware and the firmware needs to translate it. And this will have a performance reduction. To avoid that driver proactively checks whether the translation will be done in hardware or not, if not then driver try to translate inside the driver. Signed-off-by: Chaitra P B Signed-off-by: Suganath Prabu S --- drivers/scsi/mpt3sas/mpt3sas_base.c | 668 +++++++++++++++++++++++++++++- drivers/scsi/mpt3sas/mpt3sas_base.h | 44 ++- drivers/scsi/mpt3sas/mpt3sas_ctl.c | 1 + drivers/scsi/mpt3sas/mpt3sas_scsih.c | 14 +- drivers/scsi/mpt3sas/mpt3sas_warpdrive.c | 2 +- 5 files changed, 713 insertions(+), 16 deletions(-) diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.c b/drivers/scsi/mpt3sas/mpt3sas_base.c index 18039bb..d48f176 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_base.c +++ b/drivers/scsi/mpt3sas/mpt3sas_base.c @@ -59,6 +59,7 @@ #include #include #include +#include /* To get host page size per arch */ #include @@ -1347,6 +1348,547 @@ _base_build_sg(struct MPT3SAS_ADAPTER *ioc, void *psge, /* IEEE format sgls */ /** + * _base_build_nvme_prp - This function is called for NVMe end devices to build + * a native SGL (NVMe PRP). The native SGL is built starting in the first PRP + * entry of the NVMe message (PRP1). If the data buffer is small enough to be + * described entirely using PRP1, then PRP2 is not used. If needed, PRP2 is + * used to describe a larger data buffer. If the data buffer is too large to + * describe using the two PRP entriess inside the NVMe message, then PRP1 + * describes the first data memory segment, and PRP2 contains a pointer to a PRP + * list located elsewhere in memory to describe the remaining data memory + * segments. The PRP list will be contiguous. + + * The native SGL for NVMe devices is a Physical Region Page (PRP). A PRP + * consists of a list of PRP entries to describe a number of noncontigous + * physical memory segments as a single memory buffer, just as a SGL does. Note + * however, that this function is only used by the IOCTL call, so the memory + * given will be guaranteed to be contiguous. There is no need to translate + * non-contiguous SGL into a PRP in this case. All PRPs will describe + * contiguous space that is one page size each. + * + * Each NVMe message contains two PRP entries. The first (PRP1) either contains + * a PRP list pointer or a PRP element, depending upon the command. PRP2 + * contains the second PRP element if the memory being described fits within 2 + * PRP entries, or a PRP list pointer if the PRP spans more than two entries. + * + * A PRP list pointer contains the address of a PRP list, structured as a linear + * array of PRP entries. Each PRP entry in this list describes a segment of + * physical memory. + * + * Each 64-bit PRP entry comprises an address and an offset field. The address + * always points at the beginning of a 4KB physical memory page, and the offset + * describes where within that 4KB page the memory segment begins. Only the + * first element in a PRP list may contain a non-zero offest, implying that all + * memory segments following the first begin at the start of a 4KB page. + * + * Each PRP element normally describes 4KB of physical memory, with exceptions + * for the first and last elements in the list. If the memory being described + * by the list begins at a non-zero offset within the first 4KB page, then the + * first PRP element will contain a non-zero offset indicating where the region + * begins within the 4KB page. The last memory segment may end before the end + * of the 4KB segment, depending upon the overall size of the memory being + * described by the PRP list. + * + * Since PRP entries lack any indication of size, the overall data buffer length + * is used to determine where the end of the data memory buffer is located, and + * how many PRP entries are required to describe it. + * + * @ioc: per adapter object + * @smid: system request message index for getting asscociated SGL + * @nvme_encap_request: the NVMe request msg frame pointer + * @data_out_dma: physical address for WRITES + * @data_out_sz: data xfer size for WRITES + * @data_in_dma: physical address for READS + * @data_in_sz: data xfer size for READS + * + * Returns nothing. + */ +static void +_base_build_nvme_prp(struct MPT3SAS_ADAPTER *ioc, u16 smid, + Mpi26NVMeEncapsulatedRequest_t *nvme_encap_request, + dma_addr_t data_out_dma, size_t data_out_sz, dma_addr_t data_in_dma, + size_t data_in_sz) +{ + int prp_size = NVME_PRP_SIZE; + u64 *prp_entry, *prp1_entry, *prp2_entry, *prp_entry_phys; + u64 *prp_page, *prp_page_phys; + u32 offset, entry_len; + u32 page_mask_result, page_mask; + dma_addr_t paddr; + size_t length; + + /* + * Not all commands require a data transfer. If no data, just return + * without constructing any PRP. + */ + if (!data_in_sz && !data_out_sz) + return; + /* + * Set pointers to PRP1 and PRP2, which are in the NVMe command. + * PRP1 is located at a 24 byte offset from the start of the NVMe + * command. Then set the current PRP entry pointer to PRP1. + */ + prp1_entry = (u64 *)(nvme_encap_request->NVMe_Command + + NVME_CMD_PRP1_OFFSET); + prp2_entry = (u64 *)(nvme_encap_request->NVMe_Command + + NVME_CMD_PRP2_OFFSET); + prp_entry = prp1_entry; + /* + * For the PRP entries, use the specially allocated buffer of + * contiguous memory. + */ + prp_page = (u64 *)mpt3sas_base_get_pcie_sgl(ioc, smid); + prp_page_phys = (u64 *)mpt3sas_base_get_pcie_sgl_dma(ioc, smid); + + /* + * Check if we are within 1 entry of a page boundary we don't + * want our first entry to be a PRP List entry. + */ + page_mask = ioc->page_size - 1; + page_mask_result = (uintptr_t)((u8 *)prp_page + prp_size) & page_mask; + if (!page_mask_result) { + /* Bump up to next page boundary. */ + prp_page = (u64 *)((u8 *)prp_page + prp_size); + prp_page_phys = (u64 *)((u8 *)prp_page_phys + prp_size); + } + + /* + * Set PRP physical pointer, which initially points to the current PRP + * DMA memory page. + */ + prp_entry_phys = prp_page_phys; + + /* Get physical address and length of the data buffer. */ + if (data_in_sz) { + paddr = data_in_dma; + length = data_in_sz; + } else { + paddr = data_out_dma; + length = data_out_sz; + } + + /* Loop while the length is not zero. */ + while (length) { + /* + * Check if we need to put a list pointer here if we are at + * page boundary - prp_size (8 bytes). + */ + page_mask_result = + (uintptr_t)((u8 *)prp_entry_phys + prp_size) & page_mask; + if (!page_mask_result) { + /* + * This is the last entry in a PRP List, so we need to + * put a PRP list pointer here. What this does is: + * - bump the current memory pointer to the next + * address, which will be the next full page. + * - set the PRP Entry to point to that page. This + * is now the PRP List pointer. + * - bump the PRP Entry pointer the start of the + * next page. Since all of this PRP memory is + * contiguous, no need to get a new page - it's + * just the next address. + */ + prp_entry_phys++; + *prp_entry = cpu_to_le64((uintptr_t)prp_entry_phys); + prp_entry++; + } + + /* Need to handle if entry will be part of a page. */ + offset = (u32)paddr & page_mask; + entry_len = ioc->page_size - offset; + + if (prp_entry == prp1_entry) { + /* + * Must fill in the first PRP pointer (PRP1) before + * moving on. + */ + *prp1_entry = cpu_to_le64((u64)paddr); + + /* + * Now point to the second PRP entry within the + * command (PRP2). + */ + prp_entry = prp2_entry; + } else if (prp_entry == prp2_entry) { + /* + * Should the PRP2 entry be a PRP List pointer or just + * a regular PRP pointer? If there is more than one + * more page of data, must use a PRP List pointer. + */ + if (length > ioc->page_size) { + /* + * PRP2 will contain a PRP List pointer because + * more PRP's are needed with this command. The + * list will start at the beginning of the + * contiguous buffer. + */ + *prp2_entry = + cpu_to_le64((uintptr_t)prp_entry_phys); + + /* + * The next PRP Entry will be the start of the + * first PRP List. + */ + prp_entry = prp_page; + } else { + /* + * After this, the PRP Entries are complete. + * This command uses 2 PRP's and no PRP list. + */ + *prp2_entry = cpu_to_le64((u64)paddr); + } + } else { + /* + * Put entry in list and bump the addresses. + * + * After PRP1 and PRP2 are filled in, this will fill in + * all remaining PRP entries in a PRP List, one per + * each time through the loop. + */ + *prp_entry = cpu_to_le64((u64)paddr); + prp_entry++; + prp_entry_phys++; + } + + /* + * Bump the phys address of the command's data buffer by the + * entry_len. + */ + paddr += entry_len; + + /* Decrement length accounting for last partial page. */ + if (entry_len > length) + length = 0; + else + length -= entry_len; + } +} + +u32 base_mod64(u64 dividend, u32 divisor) +{ + u64 d; + u32 remainder; + + if (!divisor) + pr_err(KERN_ERR "mpt3sas : DIVISOR is zero, in div fn\n"); + d = dividend; + remainder = do_div(d, divisor); + return remainder; +} + +/** + * base_make_prp_nvme - + * Prepare PRPs(Physical Region Page)- SGLs specific to NVMe drives only + * + * @ioc: per adapter object + * @scmd: SCSI command from the mid-layer + * @mpi_request: mpi request + * @smid: msg Index + * @sge_count: scatter gather element count. + * + * Returns: true: PRPs are built + * false: IEEE SGLs needs to be built + */ +void +base_make_prp_nvme(struct MPT3SAS_ADAPTER *ioc, + struct scsi_cmnd *scmd, + Mpi25SCSIIORequest_t *mpi_request, + u16 smid, int sge_count) +{ + int sge_len, offset, num_prp_in_chain = 0; + Mpi25IeeeSgeChain64_t *main_chain_element, *ptr_first_sgl; + u64 *curr_buff; + dma_addr_t msg_phys; + u64 sge_addr; + u32 page_mask, page_mask_result; + struct scatterlist *sg_scmd; + u32 first_prp_len; + int data_len = scsi_bufflen(scmd); + u32 nvme_pg_size; + + nvme_pg_size = max_t(u32, ioc->page_size, NVME_PRP_PAGE_SIZE); + /* + * Nvme has a very convoluted prp format. One prp is required + * for each page or partial page. Driver need to split up OS sg_list + * entries if it is longer than one page or cross a page + * boundary. Driver also have to insert a PRP list pointer entry as + * the last entry in each physical page of the PRP list. + * + * NOTE: The first PRP "entry" is actually placed in the first + * SGL entry in the main message as IEEE 64 format. The 2nd + * entry in the main message is the chain element, and the rest + * of the PRP entries are built in the contiguous pcie buffer. + */ + page_mask = nvme_pg_size - 1; + + /* + * Native SGL is needed. + * Put a chain element in main message frame that points to the first + * chain buffer. + * + * NOTE: The ChainOffset field must be 0 when using a chain pointer to + * a native SGL. + */ + + /* Set main message chain element pointer */ + main_chain_element = (pMpi25IeeeSgeChain64_t)&mpi_request->SGL; + /* + * For NVMe the chain element needs to be the 2nd SG entry in the main + * message. + */ + main_chain_element = (Mpi25IeeeSgeChain64_t *) + ((u8 *)main_chain_element + sizeof(MPI25_IEEE_SGE_CHAIN64)); + + /* + * For the PRP entries, use the specially allocated buffer of + * contiguous memory. Normal chain buffers can't be used + * because each chain buffer would need to be the size of an OS + * page (4k). + */ + curr_buff = mpt3sas_base_get_pcie_sgl(ioc, smid); + msg_phys = (dma_addr_t)mpt3sas_base_get_pcie_sgl_dma(ioc, smid); + + main_chain_element->Address = cpu_to_le64(msg_phys); + main_chain_element->NextChainOffset = 0; + main_chain_element->Flags = MPI2_IEEE_SGE_FLAGS_CHAIN_ELEMENT | + MPI2_IEEE_SGE_FLAGS_SYSTEM_ADDR | + MPI26_IEEE_SGE_FLAGS_NSF_NVME_PRP; + + /* Build first prp, sge need not to be page aligned*/ + ptr_first_sgl = (pMpi25IeeeSgeChain64_t)&mpi_request->SGL; + sg_scmd = scsi_sglist(scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + + offset = (u32)(sge_addr & page_mask); + first_prp_len = nvme_pg_size - offset; + + ptr_first_sgl->Address = cpu_to_le64(sge_addr); + ptr_first_sgl->Length = cpu_to_le32(first_prp_len); + + data_len -= first_prp_len; + + if (sge_len > first_prp_len) { + sge_addr += first_prp_len; + sge_len -= first_prp_len; + } else if (sge_len == first_prp_len) { + sg_scmd = sg_next(sg_scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + } + + for (;;) { + offset = (u32)(sge_addr & page_mask); + + /* Put PRP pointer due to page boundary*/ + page_mask_result = (uintptr_t)(curr_buff + 1) & page_mask; + if (unlikely(!page_mask_result)) { + scmd_printk(KERN_NOTICE, + scmd, "page boundary curr_buff: 0x%p\n", + curr_buff); + msg_phys += 8; + *curr_buff = cpu_to_le64(msg_phys); + curr_buff++; + num_prp_in_chain++; + } + + *curr_buff = cpu_to_le64(sge_addr); + curr_buff++; + msg_phys += 8; + num_prp_in_chain++; + + sge_addr += nvme_pg_size; + sge_len -= nvme_pg_size; + data_len -= nvme_pg_size; + + if (data_len <= 0) + break; + + if (sge_len > 0) + continue; + + sg_scmd = sg_next(sg_scmd); + sge_addr = sg_dma_address(sg_scmd); + sge_len = sg_dma_len(sg_scmd); + } + + main_chain_element->Length = + cpu_to_le32(num_prp_in_chain * sizeof(u64)); + return; +} + +static bool +base_is_prp_possible(struct MPT3SAS_ADAPTER *ioc, + struct _pcie_device *pcie_device, struct scsi_cmnd *scmd, int sge_count) +{ + u32 i; + u32 data_length = 0; + struct scatterlist *sg_scmd; + bool build_prp = false; + u32 nvme_pg_size, page_mask; + u32 first_page_offset, first_page_data_size, end_residual; + u64 *msg_phys; + + nvme_pg_size = max_t(u32, ioc->page_size, + NVME_PRP_PAGE_SIZE); + data_length = cpu_to_le32(scsi_bufflen(scmd)); + sg_scmd = scsi_sglist(scmd); + + /* Create page_mask (to get offset within page) */ + page_mask = ioc->page_size - 1; + + /* + * For NVMe, check if the data transfer length exceeds the MDTS (Maximum + * Data Transfer Size) for this device. If so, just return 1 so a + * normal IEEE SGL is built. F/W will break up the I/O into multiple + * I/O's. [nvme_mdts = 0 means unlimited] + */ + + if ((pcie_device->nvme_mdts > 0) && + (data_length > pcie_device->nvme_mdts)) + return build_prp; + + /* + ** NVMe uses one PRP for each page (or part of a page) + ** look at the data length - if 4 pages or less then IEEE is OK + ** if > 5 pages then we need to build a native SGL + ** if > 4 and <= 5 pages, then check physical address of 1st SG entry + ** if this first size in the page is >= the residual beyond 4 pages + ** then use IEEE, otherwise use native SGL + **/ + + if (data_length > (nvme_pg_size * 5)) { + build_prp = true; + } else if ((data_length > (nvme_pg_size * 4)) && + (data_length <= (nvme_pg_size * 5))) { + msg_phys = (u64 *) sg_dma_address(sg_scmd); + first_page_offset = ((u32)(u64)msg_phys & page_mask); + first_page_data_size = ioc->page_size - + first_page_offset; + end_residual = data_length % ioc->page_size; + /* + ** If offset into first page pushes the end of + ** the data beyond end of the 5th page, we need + ** the extra PRP list + **/ + if (first_page_data_size < end_residual) + build_prp = true; + + /* check if 1st SG entry size is < residual beyond 4 pages */ + if (sg_dma_len(sg_scmd) < (data_length - (nvme_pg_size * 4))) + build_prp = true; + } + + /* + ** Below code detects gaps/holes in IO data buffers. + ** What does holes/gaps mean? + ** Any SGE except first one in a SGL starts at non NVME page size + ** aligned address OR Any SGE except last one in a SGL ends at + ** non NVME page size boundary. + ** + ** Driver has already informed block layer by setting boundary rules + ** for bio merging done at NVME page size boundary calling kernel API + ** blk_queue_virt_boundary inside slave_config. + ** Still there is possibility of IO coming with holes to driver because + ** of IO merging done by IO scheduler. + ** + ** With SCSI BLK MQ enabled, there will be no IO with holes as there + ** is no IO scheduling so no IO merging. + ** + ** With SCSI BLK MQ disabled, IO scheduler may attempt to merge IOs and + ** then sending IOs with holes. + ** + ** Though driver can request block layer to disable IO merging by + ** calling queue_flag_set_unlocked(QUEUE_FLAG_NOMERGES, + ** sdev->request_queue) but user may tune sysfs parameter- nomerges + ** again to 0 or 1. + ** + ** If in future IO scheduling is enabled with SCSI BLK MQ, + ** this algorithm to detect holes will be required in driver + ** for SCSI BLK MQ enabled case as well. + ** + **/ + scsi_for_each_sg(scmd, sg_scmd, sge_count, i) { + if ((i != 0) && (i != (sge_count - 1))) { + if (base_mod64(sg_dma_len(sg_scmd), nvme_pg_size) || + base_mod64(sg_dma_address(sg_scmd), + nvme_pg_size)) { + build_prp = false; + break; + } + } + + if ((sge_count > 1) && (i == 0)) { + if ((base_mod64((sg_dma_address(sg_scmd) + + sg_dma_len(sg_scmd)), nvme_pg_size))) { + build_prp = false; + break; + } + } + + if ((sge_count > 1) && (i == (sge_count - 1))) { + if (base_mod64(sg_dma_address(sg_scmd), nvme_pg_size)) { + build_prp = false; + break; + } + } + } + return build_prp; +} + +/** + * _base_check_pcie_native_sgl - This function is called for PCIe end devices to + * determine if the driver needs to build a native SGL. If so, that native + * SGL is built in the special contiguous buffers allocated especially for + * PCIe SGL creation. If the driver will not build a native SGL, return + * TRUE and a normal IEEE SGL will be built. Currently this routine + * supports NVMe. + * @ioc: per adapter object + * @mpi_request: mf request pointer + * @smid: system request message index + * @scmd: scsi command + * @pcie_device: points to the PCIe device's info + * + * Returns 0 if native SGL was built, 1 if no SGL was built + */ +static int +_base_check_pcie_native_sgl(struct MPT3SAS_ADAPTER *ioc, + Mpi25SCSIIORequest_t *mpi_request, u16 smid, struct scsi_cmnd *scmd, + struct _pcie_device *pcie_device) +{ + struct scatterlist *sg_scmd; + int sges_left; + + /* Get the SG list pointer and info. */ + sg_scmd = scsi_sglist(scmd); + sges_left = scsi_dma_map(scmd); + if (sges_left < 0) { + sdev_printk(KERN_ERR, scmd->device, + "scsi_dma_map failed: request for %d bytes!\n", + scsi_bufflen(scmd)); + return 1; + } + + /* Check if we need to build a native SG list. */ + if (base_is_prp_possible(ioc, pcie_device, + scmd, sges_left) == 0) { + /* We built a native SG list, just return. */ + goto out; + } + + /* + * Build native NVMe PRP. + */ + base_make_prp_nvme(ioc, scmd, mpi_request, + smid, sges_left); + + return 0; +out: + scsi_dma_unmap(scmd); + return 1; +} + +/** * _base_add_sg_single_ieee - add sg element for IEEE format * @paddr: virtual address for SGE * @flags: SGE flags @@ -1391,9 +1933,11 @@ _base_build_zero_len_sge_ieee(struct MPT3SAS_ADAPTER *ioc, void *paddr) /** * _base_build_sg_scmd - main sg creation routine + * pcie_device is unused here! * @ioc: per adapter object * @scmd: scsi command * @smid: system request message index + * @unused: unused pcie_device pointer * Context: none. * * The main routine that builds scatter gather table from a given @@ -1403,7 +1947,7 @@ _base_build_zero_len_sge_ieee(struct MPT3SAS_ADAPTER *ioc, void *paddr) */ static int _base_build_sg_scmd(struct MPT3SAS_ADAPTER *ioc, - struct scsi_cmnd *scmd, u16 smid) + struct scsi_cmnd *scmd, u16 smid, struct _pcie_device *unused) { Mpi2SCSIIORequest_t *mpi_request; dma_addr_t chain_dma; @@ -1537,6 +2081,8 @@ _base_build_sg_scmd(struct MPT3SAS_ADAPTER *ioc, * @ioc: per adapter object * @scmd: scsi command * @smid: system request message index + * @pcie_device: Pointer to pcie_device. If set, the pcie native sgl will be + * constructed on need. * Context: none. * * The main routine that builds scatter gather table from a given @@ -1546,9 +2092,9 @@ _base_build_sg_scmd(struct MPT3SAS_ADAPTER *ioc, */ static int _base_build_sg_scmd_ieee(struct MPT3SAS_ADAPTER *ioc, - struct scsi_cmnd *scmd, u16 smid) + struct scsi_cmnd *scmd, u16 smid, struct _pcie_device *pcie_device) { - Mpi2SCSIIORequest_t *mpi_request; + Mpi25SCSIIORequest_t *mpi_request; dma_addr_t chain_dma; struct scatterlist *sg_scmd; void *sg_local, *chain; @@ -1571,6 +2117,13 @@ _base_build_sg_scmd_ieee(struct MPT3SAS_ADAPTER *ioc, chain_sgl_flags = MPI2_IEEE_SGE_FLAGS_CHAIN_ELEMENT | MPI2_IEEE_SGE_FLAGS_SYSTEM_ADDR; + /* Check if we need to build a native SG list. */ + if ((pcie_device) && (_base_check_pcie_native_sgl(ioc, mpi_request, + smid, scmd, pcie_device) == 0)) { + /* We built a native SG list, just return. */ + return 0; + } + sg_scmd = scsi_sglist(scmd); sges_left = scsi_dma_map(scmd); if (sges_left < 0) { @@ -1582,12 +2135,12 @@ _base_build_sg_scmd_ieee(struct MPT3SAS_ADAPTER *ioc, sg_local = &mpi_request->SGL; sges_in_segment = (ioc->request_sz - - offsetof(Mpi2SCSIIORequest_t, SGL))/ioc->sge_size_ieee; + offsetof(Mpi25SCSIIORequest_t, SGL))/ioc->sge_size_ieee; if (sges_left <= sges_in_segment) goto fill_in_last_segment; mpi_request->ChainOffset = (sges_in_segment - 1 /* chain element */) + - (offsetof(Mpi2SCSIIORequest_t, SGL)/ioc->sge_size_ieee); + (offsetof(Mpi25SCSIIORequest_t, SGL)/ioc->sge_size_ieee); /* fill in main message segment when there is a chain following */ while (sges_in_segment > 1) { @@ -2267,6 +2820,33 @@ mpt3sas_base_get_sense_buffer_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid) } /** + * mpt3sas_base_get_pcie_sgl - obtain a PCIe SGL virt addr + * @ioc: per adapter object + * @smid: system request message index + * + * Returns virt pointer to a PCIe SGL. + */ +void * +mpt3sas_base_get_pcie_sgl(struct MPT3SAS_ADAPTER *ioc, u16 smid) +{ + return (void *)(ioc->scsi_lookup[smid - 1].pcie_sg_list.pcie_sgl); +} + +/** + * mpt3sas_base_get_pcie_sgl_dma - obtain a PCIe SGL dma addr + * @ioc: per adapter object + * @smid: system request message index + * + * Returns phys pointer to the address of the PCIe buffer. + */ +void * +mpt3sas_base_get_pcie_sgl_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid) +{ + return (void *)(uintptr_t) + (ioc->scsi_lookup[smid - 1].pcie_sg_list.pcie_sgl_dma); +} + +/** * mpt3sas_base_get_reply_virt_addr - obtain reply frames virt address * @ioc: per adapter object * @phys_addr: lower 32 physical addr of the reply @@ -2945,6 +3525,11 @@ _base_display_ioc_capabilities(struct MPT3SAS_ADAPTER *ioc) _base_display_OEMs_branding(ioc); + if (ioc->facts.ProtocolFlags & MPI2_IOCFACTS_PROTOCOL_NVME_DEVICES) { + pr_info("%sNVMe", i ? "," : ""); + i++; + } + pr_info(MPT3SAS_FMT "Protocol=(", ioc->name); if (ioc->facts.ProtocolFlags & MPI2_IOCFACTS_PROTOCOL_SCSI_INITIATOR) { @@ -3249,6 +3834,17 @@ _base_release_memory_pools(struct MPT3SAS_ADAPTER *ioc) kfree(ioc->reply_post); } + if (ioc->pcie_sgl_dma_pool) { + for (i = 0; i < ioc->scsiio_depth; i++) { + if (ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl) + pci_pool_free(ioc->pcie_sgl_dma_pool, + ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl, + ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl_dma); + } + if (ioc->pcie_sgl_dma_pool) + pci_pool_destroy(ioc->pcie_sgl_dma_pool); + } + if (ioc->config_page) { dexitprintk(ioc, pr_info(MPT3SAS_FMT "config_page(0x%p): free\n", ioc->name, @@ -3291,7 +3887,7 @@ _base_allocate_memory_pools(struct MPT3SAS_ADAPTER *ioc) u16 chains_needed_per_io; u32 sz, total_sz, reply_post_free_sz; u32 retry_sz; - u16 max_request_credit; + u16 max_request_credit, nvme_blocks_needed; unsigned short sg_tablesize; u16 sge_size; int i; @@ -3627,7 +4223,52 @@ _base_allocate_memory_pools(struct MPT3SAS_ADAPTER *ioc) "internal(0x%p): depth(%d), start smid(%d)\n", ioc->name, ioc->internal, ioc->internal_depth, ioc->internal_smid)); + /* + * The number of NVMe page sized blocks needed is: + * (((sg_tablesize * 8) - 1) / (page_size - 8)) + 1 + * ((sg_tablesize * 8) - 1) is the max PRP's minus the first PRP entry + * that is placed in the main message frame. 8 is the size of each PRP + * entry or PRP list pointer entry. 8 is subtracted from page_size + * because of the PRP list pointer entry at the end of a page, so this + * is not counted as a PRP entry. The 1 added page is a round up. + * + * To avoid allocation failures due to the amount of memory that could + * be required for NVMe PRP's, only each set of NVMe blocks will be + * contiguous, so a new set is allocated for each possible I/O. + */ + if (ioc->facts.ProtocolFlags & MPI2_IOCFACTS_PROTOCOL_NVME_DEVICES) { + nvme_blocks_needed = + (ioc->shost->sg_tablesize * NVME_PRP_SIZE) - 1; + nvme_blocks_needed /= (ioc->page_size - NVME_PRP_SIZE); + nvme_blocks_needed++; + + sz = nvme_blocks_needed * ioc->page_size; + ioc->pcie_sgl_dma_pool = + pci_pool_create("PCIe SGL pool", ioc->pdev, sz, 16, 0); + if (!ioc->pcie_sgl_dma_pool) { + pr_info(MPT3SAS_FMT + "PCIe SGL pool: pci_pool_create failed\n", + ioc->name); + goto out; + } + for (i = 0; i < ioc->scsiio_depth; i++) { + ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl = + pci_pool_alloc(ioc->pcie_sgl_dma_pool, + GFP_KERNEL, + &ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl_dma); + if (!ioc->scsi_lookup[i].pcie_sg_list.pcie_sgl) { + pr_info(MPT3SAS_FMT + "PCIe SGL pool: pci_pool_alloc failed\n", + ioc->name); + goto out; + } + } + dinitprintk(ioc, pr_info(MPT3SAS_FMT "PCIe sgl pool depth(%d), " + "element_size(%d), pool_size(%d kB)\n", ioc->name, + ioc->scsiio_depth, sz, (sz * ioc->scsiio_depth)/1024)); + total_sz += sz * ioc->scsiio_depth; + } /* sense buffers, 4 byte align */ sz = ioc->scsiio_depth * SCSI_SENSE_BUFFERSIZE; ioc->sense_dma_pool = pci_pool_create("sense pool", ioc->pdev, sz, 4, @@ -4472,6 +5113,19 @@ _base_get_ioc_facts(struct MPT3SAS_ADAPTER *ioc) le16_to_cpu(mpi_reply.HighPriorityCredit); facts->ReplyFrameSize = mpi_reply.ReplyFrameSize; facts->MaxDevHandle = le16_to_cpu(mpi_reply.MaxDevHandle); + facts->CurrentHostPageSize = mpi_reply.CurrentHostPageSize; + + /* + * Get the Page Size from IOC Facts. If it's 0, default to 4k. + */ + ioc->page_size = 1 << facts->CurrentHostPageSize; + if (ioc->page_size == 1) { + pr_info(MPT3SAS_FMT "CurrentHostPageSize is 0: Setting " + "default host page size to 4k\n", ioc->name); + ioc->page_size = 1 << MPT3SAS_HOST_PAGE_SIZE_4K; + } + dinitprintk(ioc, pr_info(MPT3SAS_FMT "CurrentHostPageSize(%d)\n", + ioc->name, facts->CurrentHostPageSize)); dinitprintk(ioc, pr_info(MPT3SAS_FMT "hba queue depth(%d), max chains per io(%d)\n", @@ -4511,6 +5165,7 @@ _base_send_ioc_init(struct MPT3SAS_ADAPTER *ioc) mpi_request.VP_ID = 0; mpi_request.MsgVersion = cpu_to_le16(ioc->hba_mpi_version_belonged); mpi_request.HeaderVersion = cpu_to_le16(MPI2_HEADER_VERSION); + mpi_request.HostPageSize = MPT3SAS_HOST_PAGE_SIZE_4K; if (_base_is_controller_msix_enabled(ioc)) mpi_request.HostMSIxVectors = ioc->reply_queue_count; @@ -5379,6 +6034,7 @@ mpt3sas_base_attach(struct MPT3SAS_ADAPTER *ioc) */ ioc->build_sg_scmd = &_base_build_sg_scmd_ieee; ioc->build_sg = &_base_build_sg_ieee; + ioc->build_nvme_prp = &_base_build_nvme_prp; ioc->build_zero_len_sge = &_base_build_zero_len_sge_ieee; ioc->sge_size_ieee = sizeof(Mpi2IeeeSgeSimple64_t); diff --git a/drivers/scsi/mpt3sas/mpt3sas_base.h b/drivers/scsi/mpt3sas/mpt3sas_base.h index c522057..4caa91e 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_base.h +++ b/drivers/scsi/mpt3sas/mpt3sas_base.h @@ -54,6 +54,7 @@ #include "mpi/mpi2_raid.h" #include "mpi/mpi2_tool.h" #include "mpi/mpi2_sas.h" +#include "mpi/mpi2_pci.h" #include #include @@ -113,6 +114,7 @@ #define MPT3SAS_RAID_QUEUE_DEPTH 128 #define MPT3SAS_RAID_MAX_SECTORS 8192 +#define MPT3SAS_HOST_PAGE_SIZE_4K 12 #define MPT_NAME_LENGTH 32 /* generic length of strings */ #define MPT_STRING_LENGTH 64 @@ -131,6 +133,15 @@ #define DEFAULT_NUM_FWCHAIN_ELEMTS 8 /* + * NVMe defines + */ +#define NVME_PRP_SIZE 8 /* PRP size */ +#define NVME_CMD_PRP1_OFFSET 24 /* PRP1 offset in NVMe cmd */ +#define NVME_CMD_PRP2_OFFSET 32 /* PRP2 offset in NVMe cmd */ +#define NVME_ERROR_RESPONSE_SIZE 16 /* Max NVME Error Response */ +#define NVME_PRP_PAGE_SIZE 4096 /* Page size */ + +/* * reset phases */ #define MPT3_IOC_PRE_RESET 1 /* prior to host reset */ @@ -731,6 +742,16 @@ enum reset_type { }; /** + * struct pcie_sg_list - PCIe SGL buffer (contiguous per I/O) + * @pcie_sgl: PCIe native SGL for NVMe devices + * @pcie_sgl_dma: physical address + */ +struct pcie_sg_list { + void *pcie_sgl; + dma_addr_t pcie_sgl_dma; +}; + +/** * struct chain_tracker - firmware chain tracker * @chain_buffer: chain buffer * @chain_buffer_dma: physical address @@ -756,6 +777,7 @@ struct scsiio_tracker { struct scsi_cmnd *scmd; u8 cb_idx; u8 direct_io; + struct pcie_sg_list pcie_sg_list; struct list_head chain_list; struct list_head tracker_list; u16 msix_io; @@ -829,13 +851,19 @@ typedef void (*MPT_ADD_SGE)(void *paddr, u32 flags_length, dma_addr_t dma_addr); /* SAS3.0 support */ typedef int (*MPT_BUILD_SG_SCMD)(struct MPT3SAS_ADAPTER *ioc, - struct scsi_cmnd *scmd, u16 smid); + struct scsi_cmnd *scmd, u16 smid, struct _pcie_device *pcie_device); typedef void (*MPT_BUILD_SG)(struct MPT3SAS_ADAPTER *ioc, void *psge, dma_addr_t data_out_dma, size_t data_out_sz, dma_addr_t data_in_dma, size_t data_in_sz); typedef void (*MPT_BUILD_ZERO_LEN_SGE)(struct MPT3SAS_ADAPTER *ioc, void *paddr); +/* SAS3.5 support */ +typedef void (*NVME_BUILD_PRP)(struct MPT3SAS_ADAPTER *ioc, u16 smid, + Mpi26NVMeEncapsulatedRequest_t *nvme_encap_request, + dma_addr_t data_out_dma, size_t data_out_sz, dma_addr_t data_in_dma, + size_t data_in_sz); + /* To support atomic and non atomic descriptors*/ typedef void (*PUT_SMID_IO_FP_HIP) (struct MPT3SAS_ADAPTER *ioc, u16 smid, u16 funcdep); @@ -878,6 +906,7 @@ struct mpt3sas_facts { u16 MaxDevHandle; u16 MaxPersistentEntries; u16 MinDevHandle; + u8 CurrentHostPageSize; }; struct mpt3sas_port_facts { @@ -1149,6 +1178,9 @@ struct MPT3SAS_ADAPTER { MPT_BUILD_SG build_sg_mpi; MPT_BUILD_ZERO_LEN_SGE build_zero_len_sge_mpi; + /* function ptr for NVMe PRP elements only */ + NVME_BUILD_PRP build_nvme_prp; + /* event log */ u32 event_type[MPI2_EVENT_NOTIFY_EVENTMASK_WORDS]; u32 event_context; @@ -1217,6 +1249,11 @@ struct MPT3SAS_ADAPTER { int pending_io_count; wait_queue_head_t reset_wq; + /* PCIe SGL */ + struct dma_pool *pcie_sgl_dma_pool; + /* Host Page Size */ + u32 page_size; + /* chain */ struct chain_tracker *chain_lookup; struct list_head free_chain_list; @@ -1350,7 +1387,8 @@ void *mpt3sas_base_get_msg_frame(struct MPT3SAS_ADAPTER *ioc, u16 smid); void *mpt3sas_base_get_sense_buffer(struct MPT3SAS_ADAPTER *ioc, u16 smid); __le32 mpt3sas_base_get_sense_buffer_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid); - +void *mpt3sas_base_get_pcie_sgl(struct MPT3SAS_ADAPTER *ioc, u16 smid); +void *mpt3sas_base_get_pcie_sgl_dma(struct MPT3SAS_ADAPTER *ioc, u16 smid); void mpt3sas_base_sync_reply_irqs(struct MPT3SAS_ADAPTER *ioc); /* hi-priority queue */ @@ -1564,7 +1602,7 @@ void mpt3sas_scsi_direct_io_set(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 direct_io); void mpt3sas_setup_direct_io(struct MPT3SAS_ADAPTER *ioc, struct scsi_cmnd *scmd, - struct _raid_device *raid_device, Mpi2SCSIIORequest_t *mpi_request, + struct _raid_device *raid_device, Mpi25SCSIIORequest_t *mpi_request, u16 smid); /* NCQ Prio Handling Check */ diff --git a/drivers/scsi/mpt3sas/mpt3sas_ctl.c b/drivers/scsi/mpt3sas/mpt3sas_ctl.c index bdffb69..0c18831 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_ctl.c +++ b/drivers/scsi/mpt3sas/mpt3sas_ctl.c @@ -299,6 +299,7 @@ mpt3sas_ctl_done(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 msix_index, } } } + _ctl_display_some_debug(ioc, smid, "ctl_done", mpi_reply); ioc->ctl_cmds.status &= ~MPT3_CMD_PENDING; complete(&ioc->ctl_cmds.done); diff --git a/drivers/scsi/mpt3sas/mpt3sas_scsih.c b/drivers/scsi/mpt3sas/mpt3sas_scsih.c index 4b6b0fa..3e6dd4b 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_scsih.c +++ b/drivers/scsi/mpt3sas/mpt3sas_scsih.c @@ -4244,7 +4244,7 @@ _scsih_flush_running_cmds(struct MPT3SAS_ADAPTER *ioc) */ static void _scsih_setup_eedp(struct MPT3SAS_ADAPTER *ioc, struct scsi_cmnd *scmd, - Mpi2SCSIIORequest_t *mpi_request) + Mpi25SCSIIORequest_t *mpi_request) { u16 eedp_flags; unsigned char prot_op = scsi_get_prot_op(scmd); @@ -4347,7 +4347,8 @@ scsih_qcmd(struct Scsi_Host *shost, struct scsi_cmnd *scmd) struct _raid_device *raid_device; struct request *rq = scmd->request; int class; - Mpi2SCSIIORequest_t *mpi_request; + Mpi25SCSIIORequest_t *mpi_request; + struct _pcie_device *pcie_device = NULL; u32 mpi_control; u16 smid; u16 handle; @@ -4435,7 +4436,7 @@ scsih_qcmd(struct Scsi_Host *shost, struct scsi_cmnd *scmd) goto out; } mpi_request = mpt3sas_base_get_msg_frame(ioc, smid); - memset(mpi_request, 0, sizeof(Mpi2SCSIIORequest_t)); + memset(mpi_request, 0, ioc->request_sz); _scsih_setup_eedp(ioc, scmd, mpi_request); if (scmd->cmd_len == 32) @@ -4454,13 +4455,14 @@ scsih_qcmd(struct Scsi_Host *shost, struct scsi_cmnd *scmd) mpi_request->SenseBufferLength = SCSI_SENSE_BUFFERSIZE; mpi_request->SenseBufferLowAddress = mpt3sas_base_get_sense_buffer_dma(ioc, smid); - mpi_request->SGLOffset0 = offsetof(Mpi2SCSIIORequest_t, SGL) / 4; + mpi_request->SGLOffset0 = offsetof(Mpi25SCSIIORequest_t, SGL) / 4; int_to_scsilun(sas_device_priv_data->lun, (struct scsi_lun *) mpi_request->LUN); memcpy(mpi_request->CDB.CDB32, scmd->cmnd, scmd->cmd_len); if (mpi_request->DataLength) { - if (ioc->build_sg_scmd(ioc, scmd, smid)) { + pcie_device = sas_target_priv_data->pcie_dev; + if (ioc->build_sg_scmd(ioc, scmd, smid, pcie_device)) { mpt3sas_base_free_smid(ioc, smid); goto out; } @@ -4931,7 +4933,7 @@ out_unlock: static u8 _scsih_io_done(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 msix_index, u32 reply) { - Mpi2SCSIIORequest_t *mpi_request; + Mpi25SCSIIORequest_t *mpi_request; Mpi2SCSIIOReply_t *mpi_reply; struct scsi_cmnd *scmd; u16 ioc_status; diff --git a/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c b/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c index 540bd50..ced7d9f 100644 --- a/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c +++ b/drivers/scsi/mpt3sas/mpt3sas_warpdrive.c @@ -299,7 +299,7 @@ mpt3sas_scsi_direct_io_set(struct MPT3SAS_ADAPTER *ioc, u16 smid, u8 direct_io) */ void mpt3sas_setup_direct_io(struct MPT3SAS_ADAPTER *ioc, struct scsi_cmnd *scmd, - struct _raid_device *raid_device, Mpi2SCSIIORequest_t *mpi_request, + struct _raid_device *raid_device, Mpi25SCSIIORequest_t *mpi_request, u16 smid) { sector_t v_lba, p_lba, stripe_off, column, io_size; -- 1.7.1