Re: [PATCH v8 15/45] powerpc/powernv/ioda1: Introduce PNV_IODA1_DMA32_SEGSIZE

[Gavin Shan <gwshan@linux.vnet.ibm.com>]

On Thu, Apr 14, 2016 at 01:36:33PM +1000, Alexey Kardashevskiy wrote:
>On 04/14/2016 09:54 AM, Gavin Shan wrote:
>>On Wed, Apr 13, 2016 at 06:29:42PM +1000, Alexey Kardashevskiy wrote:
>>>On 02/17/2016 02:43 PM, Gavin Shan wrote:
>>>>Currently, there is one macro (TCE32_TABLE_SIZE) representing the
>>>>TCE table size for one DMA32 segment. The constant representing
>>>>the DMA32 segment size (1 << 28) is still used in the code.
>>>>
>>>>This defines PNV_IODA1_DMA32_SEGSIZE representing one DMA32
>>>>segment size. the TCE table size can be calcualted when the page
>>>
>>>s/calcualted/calculated/
>>>
>>>
>>>>has fixed 4KB size. So all the related calculation depends on one
>>>>macro (PNV_IODA1_DMA32_SEGSIZE). No logical changes introduced.
>>>
>>>Please move PNV_IODA1_DMA32_SEGSIZE where TCE32_TABLE_SIZE was.
>>>
>>>
>>>>
>>>>Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
>>>>---
>>>>  arch/powerpc/platforms/powernv/pci-ioda.c | 30 +++++++++++++++++-------------
>>>>  arch/powerpc/platforms/powernv/pci.h      |  1 +
>>>>  2 files changed, 18 insertions(+), 13 deletions(-)
>>>>
>>>>diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
>>>>index d18b95e..e60cff6 100644
>>>>--- a/arch/powerpc/platforms/powernv/pci-ioda.c
>>>>+++ b/arch/powerpc/platforms/powernv/pci-ioda.c
>>>>@@ -48,9 +48,6 @@
>>>>  #include "powernv.h"
>>>>  #include "pci.h"
>>>>
>>>>-/* 256M DMA window, 4K TCE pages, 8 bytes TCE */
>>>>-#define TCE32_TABLE_SIZE	((0x10000000 / 0x1000) * 8)
>>>>-
>>>>  #define POWERNV_IOMMU_DEFAULT_LEVELS	1
>>>>  #define POWERNV_IOMMU_MAX_LEVELS	5
>>>>
>>>>@@ -2034,7 +2031,7 @@ static void pnv_pci_ioda1_setup_dma_pe(struct pnv_phb *phb,
>>>>
>>>>  	struct page *tce_mem = NULL;
>>>>  	struct iommu_table *tbl;
>>>>-	unsigned int i;
>>>>+	unsigned int tce32_segsz, i;
>>>
>>>
>>>PNV_IODA1_DMA32_SEGSIZE is a segment size in bytes. The name @tce32_segsz
>>>also suggests that it is a segment size in bytes (otherwise it would be
>>>tce32_seg_entries or something like this) but it is not, it is a number of
>>>TCE entries (arch/powerpc/kernel/iommu.c uses "entry" for these). And
>>>tce32_segsz never changes. So:
>>>
>>>const unsigned int entries = PNV_IODA1_DMA32_SEGSIZE >> (IOMMU_PAGE_SHIFT_4K
>>>- 3);
>>>
>>
>>Are you sure @tce32_segsz and equation you gave are for number of TCE entries,
>>not the size of meory required for the DMA32 segment TCE table?
>
>No, I am not :) "-3" makes it a table size in bytes, so it is rather tablesz
>then.
>

Ok. @tce32_segsz is the size of memory used for TCE entries for one segment (256MB),
not a whole TCE table. So I think @tce32_segsz is better than @tablesz from the
perspective.

>
>>
>>>>  	int64_t rc;
>>>>  	void *addr;
>>>>
>>>>@@ -2054,29 +2051,34 @@ static void pnv_pci_ioda1_setup_dma_pe(struct pnv_phb *phb,
>>>>  	/* Grab a 32-bit TCE table */
>>>>  	pe->tce32_seg = base;
>>>>  	pe_info(pe, " Setting up 32-bit TCE table at %08x..%08x\n",
>>>>-		(base << 28), ((base + segs) << 28) - 1);
>>>>+		base * PNV_IODA1_DMA32_SEGSIZE,
>>>>+		(base + segs) * PNV_IODA1_DMA32_SEGSIZE - 1);
>>>>
>>>>  	/* XXX Currently, we allocate one big contiguous table for the
>>>>  	 * TCEs. We only really need one chunk per 256M of TCE space
>>>>  	 * (ie per segment) but that's an optimization for later, it
>>>>  	 * requires some added smarts with our get/put_tce implementation
>>>>+	 *
>>>>+	 * Each TCE page is 4KB in size and each TCE entry occupies 8
>>>>+	 * bytes
>>>>  	 */
>>>>+	tce32_segsz = PNV_IODA1_DMA32_SEGSIZE >> (IOMMU_PAGE_SHIFT_4K - 3);
>>>
>>>>  	tce_mem = alloc_pages_node(phb->hose->node, GFP_KERNEL,
>>>>-				   get_order(TCE32_TABLE_SIZE * segs));
>>>>+				   get_order(tce32_segsz * segs));
>>>>  	if (!tce_mem) {
>>>>  		pe_err(pe, " Failed to allocate a 32-bit TCE memory\n");
>>>>  		goto fail;
>>>>  	}
>>>>  	addr = page_address(tce_mem);
>>>>-	memset(addr, 0, TCE32_TABLE_SIZE * segs);
>>>>+	memset(addr, 0, tce32_segsz * segs);
>>>>
>>>>  	/* Configure HW */
>>>>  	for (i = 0; i < segs; i++) {
>>>>  		rc = opal_pci_map_pe_dma_window(phb->opal_id,
>>>>  					      pe->pe_number,
>>>>  					      base + i, 1,
>>>>-					      __pa(addr) + TCE32_TABLE_SIZE * i,
>>>>-					      TCE32_TABLE_SIZE, 0x1000);
>>>>+					      __pa(addr) + tce32_segsz * i,
>>>>+					      tce32_segsz, 0x1000);
>>>
>>>
>>>As you started using IOMMU_PAGE_SHIFT_4K and you are also touching this piece
>>>of code -
>>>
>>>s/0x1000/IOMMU_PAGE_SHIFT_4K/
>>>
>>
>>Does 0x1000 is equal to IOMMU_PAGE_SHIFT_4K? I guess you probably suggested
>>to use IOMMU_PAGE_SIZE_4K instead?
>
>
>Ah, my bad, should have been IOMMU_PAGE_SIZE_4K. I'll pay more attention to
>the details, sorry.
>

No worries. Thanks for your review anyway.

>>
>>>>  		if (rc) {
>>>>  			pe_err(pe, " Failed to configure 32-bit TCE table,"
>>>>  			       " err %ld\n", rc);
>>>>@@ -2085,8 +2087,9 @@ static void pnv_pci_ioda1_setup_dma_pe(struct pnv_phb *phb,
>>>>  	}
>>>>
>>>>  	/* Setup linux iommu table */
>>>>-	pnv_pci_setup_iommu_table(tbl, addr, TCE32_TABLE_SIZE * segs,
>>>>-				  base << 28, IOMMU_PAGE_SHIFT_4K);
>>>>+	pnv_pci_setup_iommu_table(tbl, addr, tce32_segsz * segs,
>>>>+				  base * PNV_IODA1_DMA32_SEGSIZE,
>>>>+				  IOMMU_PAGE_SHIFT_4K);
>>>>
>>>>  	/* OPAL variant of P7IOC SW invalidated TCEs */
>>>>  	if (phb->ioda.tce_inval_reg)
>>>>@@ -2116,7 +2119,7 @@ static void pnv_pci_ioda1_setup_dma_pe(struct pnv_phb *phb,
>>>>  	if (pe->tce32_seg >= 0)
>>>>  		pe->tce32_seg = -1;
>>>>  	if (tce_mem)
>>>>-		__free_pages(tce_mem, get_order(TCE32_TABLE_SIZE * segs));
>>>>+		__free_pages(tce_mem, get_order(tce32_segsz * segs));
>>>>  	if (tbl) {
>>>>  		pnv_pci_unlink_table_and_group(tbl, &pe->table_group);
>>>>  		iommu_free_table(tbl, "pnv");
>>>>@@ -3445,7 +3448,8 @@ static void __init pnv_pci_init_ioda_phb(struct device_node *np,
>>>>  	mutex_init(&phb->ioda.pe_list_mutex);
>>>>
>>>>  	/* Calculate how many 32-bit TCE segments we have */
>>>>-	phb->ioda.tce32_count = phb->ioda.m32_pci_base >> 28;
>>>>+	phb->ioda.tce32_count = phb->ioda.m32_pci_base /
>>>>+				PNV_IODA1_DMA32_SEGSIZE;
>>>>
>>>>  #if 0 /* We should really do that ... */
>>>>  	rc = opal_pci_set_phb_mem_window(opal->phb_id,
>>>>diff --git a/arch/powerpc/platforms/powernv/pci.h b/arch/powerpc/platforms/powernv/pci.h
>>>>index 00539ff..1d8e775 100644
>>>>--- a/arch/powerpc/platforms/powernv/pci.h
>>>>+++ b/arch/powerpc/platforms/powernv/pci.h
>>>>@@ -84,6 +84,7 @@ struct pnv_ioda_pe {
>>>>
>>>>  #define PNV_IODA1_M64_NUM	16	/* Number of M64 BARs   */
>>>>  #define PNV_IODA1_M64_SEGS	8	/* Segments per M64 BAR */
>>>>+#define PNV_IODA1_DMA32_SEGSIZE	0x10000000
>>>>
>>>>  #define PNV_PHB_FLAG_EEH	(1 << 0)
>>>>
>>>>
>
>-- 
>Alexey
>