[PATCH v3 03/13] mpt3sas: SGL to PRP Translation for I/Os to NVMe devices

From: Suganath Prabu S
Date: Tue Aug 08 2017 - 05:05:02 EST


* 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 <chaitra.basappa@xxxxxxxxxxxx>
Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@xxxxxxxxxxxx>
---
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 <linux/time.h>
#include <linux/ktime.h>
#include <linux/kthread.h>
+#include <asm/page.h> /* To get host page size per arch */
#include <linux/aer.h>


@@ -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 <scsi/scsi.h>
#include <scsi/scsi_cmnd.h>
@@ -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