@@ -14,6 +14,8 @@
#include <linux/err.h>
#include <linux/ctype.h>
#include <linux/processor.h>
+#include <linux/dma-mapping.h>
+#include <linux/mm.h>
#include "ism.h"
@@ -292,13 +294,15 @@ static int ism_read_local_gid(struct ism_dev *ism)
static void ism_free_dmb(struct ism_dev *ism, struct ism_dmb *dmb)
{
clear_bit(dmb->sba_idx, ism->sba_bitmap);
- dma_free_coherent(&ism->pdev->dev, dmb->dmb_len,
- dmb->cpu_addr, dmb->dma_addr);
+ dma_unmap_page(&ism->pdev->dev, dmb->dma_addr, dmb->dmb_len,
+ DMA_FROM_DEVICE);
+ folio_put(virt_to_folio(dmb->cpu_addr));
}
static int ism_alloc_dmb(struct ism_dev *ism, struct ism_dmb *dmb)
{
unsigned long bit;
+ int rc;
if (PAGE_ALIGN(dmb->dmb_len) > dma_get_max_seg_size(&ism->pdev->dev))
return -EINVAL;
@@ -315,14 +319,30 @@ static int ism_alloc_dmb(struct ism_dev *ism, struct ism_dmb *dmb)
test_and_set_bit(dmb->sba_idx, ism->sba_bitmap))
return -EINVAL;
- dmb->cpu_addr = dma_alloc_coherent(&ism->pdev->dev, dmb->dmb_len,
- &dmb->dma_addr,
- GFP_KERNEL | __GFP_NOWARN |
- __GFP_NOMEMALLOC | __GFP_NORETRY);
- if (!dmb->cpu_addr)
- clear_bit(dmb->sba_idx, ism->sba_bitmap);
+ dmb->cpu_addr =
+ folio_address(folio_alloc(GFP_KERNEL | __GFP_NOWARN |
+ __GFP_NOMEMALLOC | __GFP_NORETRY,
+ get_order(dmb->dmb_len)));
- return dmb->cpu_addr ? 0 : -ENOMEM;
+ if (!dmb->cpu_addr) {
+ rc = -ENOMEM;
+ goto out_bit;
+ }
+ dmb->dma_addr = dma_map_page(&ism->pdev->dev,
+ virt_to_page(dmb->cpu_addr), 0,
+ dmb->dmb_len, DMA_FROM_DEVICE);
+ if (dma_mapping_error(&ism->pdev->dev, dmb->dma_addr)) {
+ rc = -ENOMEM;
+ goto out_free;
+ }
+
+ return 0;
+
+out_free:
+ kfree(dmb->cpu_addr);
+out_bit:
+ clear_bit(dmb->sba_idx, ism->sba_bitmap);
+ return rc;
}
int ism_register_dmb(struct ism_dev *ism, struct ism_dmb *dmb,
Since [1], dma_alloc_coherent() does not accept requests for GFP_COMP anymore, even on archs that may be able to fulfill this. Functionality that relied on the receive buffer being a compound page broke at that point: The SMC-D protocol, that utilizes the ism device driver, passes receive buffers to the splice processor in a struct splice_pipe_desc with a single entry list of struct pages. As the buffer is no longer a compound page, the splice processor now rejects requests to handle more than a page worth of data. Replace dma_alloc_coherent() and allocate a buffer with folio_alloc and create a DMA map for it with dma_map_page(). Since only receive buffers on ISM devices use DMA, qualify the mapping as FROM_DEVICE. Since ISM devices are available on arch s390, only and on that arch all DMA is coherent, there is no need to introduce and export some kind of dma_sync_to_cpu() method to be called by the SMC-D protocol layer. Analogously, replace dma_free_coherent by a two step dma_unmap_page, then folio_put to free the receive buffer. [1] https://lore.kernel.org/all/20221113163535.884299-1-hch@lst.de/ Fixes: c08004eede4b ("s390/ism: don't pass bogus GFP_ flags to dma_alloc_coherent") Signed-off-by: Gerd Bayer <gbayer@linux.ibm.com> --- changelog: v1->v2: - replaced kmalloc/kfree with folio_alloc/folio_put as suggested. --- drivers/s390/net/ism_drv.c | 38 +++++++++++++++++++++++++++++--------- 1 file changed, 29 insertions(+), 9 deletions(-)