@@ -5481,6 +5481,52 @@ _base_allocate_pcie_sgl_pool(struct MPT3SAS_ADAPTER *ioc, u32 sz)
return 0;
}
+/**
+ * _base_allocate_chain_dma_pool - Allocating DMA'able memory
+ * for chain dma pool.
+ * @ioc: Adapter object
+ * @sz: DMA Pool size
+ * @ctr: Chain tracker
+ * Return: 0 for success, non-zero for failure.
+ */
+static int
+_base_allocate_chain_dma_pool(struct MPT3SAS_ADAPTER *ioc, u32 sz)
+{
+ int i = 0, j = 0;
+ struct chain_tracker *ctr;
+
+ ioc->chain_dma_pool = dma_pool_create("chain pool", &ioc->pdev->dev,
+ ioc->chain_segment_sz, 16, 0);
+ if (!ioc->chain_dma_pool)
+ return -ENOMEM;
+
+ for (i = 0; i < ioc->scsiio_depth; i++) {
+ for (j = ioc->chains_per_prp_buffer;
+ j < ioc->chains_needed_per_io; j++) {
+ ctr = &ioc->chain_lookup[i].chains_per_smid[j];
+ ctr->chain_buffer = dma_pool_alloc(ioc->chain_dma_pool,
+ GFP_KERNEL, &ctr->chain_buffer_dma);
+ if (!ctr->chain_buffer)
+ return -EAGAIN;
+ if (!mpt3sas_check_same_4gb_region((long)
+ ctr->chain_buffer, ioc->chain_segment_sz)) {
+ ioc_err(ioc,
+ "Chain buffers are not in same 4G !!! Chain buff (0x%p) dma = (0x%llx)\n",
+ ctr->chain_buffer,
+ (unsigned long long)ctr->chain_buffer_dma);
+ ioc->use_32bit_dma = true;
+ return -EAGAIN;
+ }
+ }
+ }
+ dinitprintk(ioc, ioc_info(ioc,
+ "chain_lookup depth (%d), frame_size(%d), pool_size(%d kB)\n",
+ ioc->scsiio_depth, ioc->chain_segment_sz, ((ioc->scsiio_depth *
+ (ioc->chains_needed_per_io - ioc->chains_per_prp_buffer) *
+ ioc->chain_segment_sz))/1024));
+ return 0;
+}
+
/**
* base_alloc_rdpq_dma_pool - Allocating DMA'able memory
* for reply queues.
@@ -5578,9 +5624,8 @@ _base_allocate_memory_pools(struct MPT3SAS_ADAPTER *ioc)
u16 max_request_credit, nvme_blocks_needed;
unsigned short sg_tablesize;
u16 sge_size;
- int i, j;
+ int i;
int ret = 0, rc = 0;
- struct chain_tracker *ct;
dinitprintk(ioc, ioc_info(ioc, "%s\n", __func__));
@@ -5907,31 +5952,17 @@ _base_allocate_memory_pools(struct MPT3SAS_ADAPTER *ioc)
total_sz += sz * ioc->scsiio_depth;
}
- ioc->chain_dma_pool = dma_pool_create("chain pool", &ioc->pdev->dev,
- ioc->chain_segment_sz, 16, 0);
- if (!ioc->chain_dma_pool) {
- ioc_err(ioc, "chain_dma_pool: dma_pool_create failed\n");
- goto out;
- }
- for (i = 0; i < ioc->scsiio_depth; i++) {
- for (j = ioc->chains_per_prp_buffer;
- j < ioc->chains_needed_per_io; j++) {
- ct = &ioc->chain_lookup[i].chains_per_smid[j];
- ct->chain_buffer = dma_pool_alloc(
- ioc->chain_dma_pool, GFP_KERNEL,
- &ct->chain_buffer_dma);
- if (!ct->chain_buffer) {
- ioc_err(ioc, "chain_lookup: pci_pool_alloc failed\n");
- goto out;
- }
- }
- total_sz += ioc->chain_segment_sz;
- }
-
+ rc = _base_allocate_chain_dma_pool(ioc, ioc->chain_segment_sz);
+ if (rc == -ENOMEM)
+ return -ENOMEM;
+ else if (rc == -EAGAIN)
+ goto try_32bit_dma;
+ total_sz += ioc->chain_segment_sz * ((ioc->chains_needed_per_io -
+ ioc->chains_per_prp_buffer) * ioc->scsiio_depth);
dinitprintk(ioc,
- ioc_info(ioc, "chain pool depth(%d), frame_size(%d), pool_size(%d kB)\n",
- ioc->chain_depth, ioc->chain_segment_sz,
- (ioc->chain_depth * ioc->chain_segment_sz) / 1024));
+ ioc_info(ioc, "chain pool depth(%d), frame_size(%d), pool_size(%d kB)\n",
+ ioc->chain_depth, ioc->chain_segment_sz,
+ (ioc->chain_depth * ioc->chain_segment_sz) / 1024));
/* sense buffers, 4 byte align */
sz = ioc->scsiio_depth * SCSI_SENSE_BUFFERSIZE;
According to MPI Specification Chain buffers should not cross 4GB boundary. So while allocating Chain buffers, if any buffer crosses the 4GB boundary then, * Release the already allocated memory pools and * Reallocate them by changing the DMA coherent mask to 32 bit. Signed-off-by: Suganath Prabu S <suganath-prabu.subramani@broadcom.com> Reported-by: kernel test robot <lkp@intel.com> --- In V2 fixed inconsistent indenting warning. drivers/scsi/mpt3sas/mpt3sas_base.c | 83 ++++++++++++++++++++--------- 1 file changed, 57 insertions(+), 26 deletions(-)