Skip to contentSkip to conversation

remove dma_zalloc_coherent

GIT PULL Merged Christoph Hellwig 3 participants 5 messages Jan 12, 2019

lore
Share & tools
lore
#0 lore raw

Lets see if this works better as a pull request than the plain patches:

We've always had a weird situation around dma_zalloc_coherent. To safely support mapping the allocations to userspace major architectures like x86 and arm have always zeroed allocations from dma_alloc_coherent, but a couple other architectures were missing that zeroing either always or in corner cases. Then later we grew anothe dma_zalloc_coherent interface to explicitly request zeroing, but that just added __GFP_ZERO to the allocation flags, which for some allocators that didn't end up using the page allocator ended up being a no-op and still not zeroing the allocations.

So for this merge window I fixed up all remaining architectures to zero the memory in dma_alloc_coherent, and made dma_zalloc_coherent a no-op wrapper around dma_alloc_coherent, which fixes all of the above issues.

dma_zalloc_coherent is now pointless and can go away, and Luis helped me writing a cocchinelle script and patch series to kill it, which I think we should apply now just after -rc1 to finally settle these issue.

The following changes since commit 3bd6e94bec122a951d462c239b47954cf5f36e33:

arch: restore generic-y += shmparam.h for some architectures (2019-01-06 18:16:11 -0800)

are available in the Git repository at:

git://git.infradead.org/users/hch/dma-mapping.git tags/remove-dma_zalloc_coherent-5.0

for you to fetch changes up to dfd32cad146e3624970eee9329e99d2c6ef751b3:

dma-mapping: remove dma_zalloc_coherent() (2019-01-08 07:58:49 -0500)

----------------------------------------------------------------
remove dma_zalloc_coherent

We've always had a weird situation around dma_zalloc_coherent. To safely support mapping the allocations to userspace major architectures like x86 and arm have always zeroed allocations from dma_alloc_coherent, but a couple other architectures were missing that zeroing either always or in corner cases. Then later we grew anothe dma_zalloc_coherent interface to explicitly request zeroing, but that just added __GFP_ZERO to the allocation flags, which for some allocators that didn't end up using the page allocator ended up being a no-op and still not zeroing the allocations.

So for this merge window I fixed up all remaining architectures to zero the memory in dma_alloc_coherent, and made dma_zalloc_coherent a no-op wrapper around dma_alloc_coherent, which fixes all of the above issues.

dma_zalloc_coherent is now pointless and can go away, and Luis helped me writing a cocchinelle script and patch series to kill it, which I think we should apply now just after -rc1 to finally settle these issue.

----------------------------------------------------------------
Luis Chamberlain (3):
cross-tree: phase out dma_zalloc_coherent()
cross-tree: phase out dma_zalloc_coherent() on headers
dma-mapping: remove dma_zalloc_coherent()

 arch/mips/lantiq/xway/dma.c                        |  6 +-
 arch/powerpc/platforms/pasemi/dma_lib.c            |  2 +-
 arch/powerpc/sysdev/fsl_rmu.c                      |  7 ++-
 drivers/ata/sata_fsl.c                             |  4 +-
 drivers/atm/he.c                                   | 39 ++++++-------
 drivers/atm/idt77252.c                             | 16 ++---
 drivers/block/skd_main.c                           |  4 +-
 drivers/crypto/amcc/crypto4xx_core.c               |  6 +-
 drivers/crypto/cavium/cpt/cptpf_main.c             |  4 +-
 drivers/crypto/cavium/cpt/cptvf_main.c             |  7 ++-
 drivers/crypto/cavium/nitrox/nitrox_lib.c          |  6 +-
 drivers/crypto/ccp/ccp-dev-v5.c                    |  6 +-
 drivers/crypto/hisilicon/sec/sec_algs.c            |  4 +-
 drivers/crypto/hisilicon/sec/sec_drv.c             | 15 ++---
 drivers/crypto/ixp4xx_crypto.c                     |  6 +-
 drivers/crypto/mediatek/mtk-platform.c             | 16 ++---
 drivers/crypto/qat/qat_common/adf_admin.c          | 12 ++--
 drivers/crypto/qat/qat_common/qat_algs.c           | 24 ++++----
 drivers/crypto/qat/qat_common/qat_asym_algs.c      | 68 +++++++++++-----------
 drivers/dma/imx-sdma.c                             |  8 +--
 drivers/dma/mediatek/mtk-hsdma.c                   |  4 +-
 drivers/dma/mxs-dma.c                              |  6 +-
 drivers/dma/xgene-dma.c                            |  4 +-
 drivers/dma/xilinx/xilinx_dma.c                    | 14 ++---
 drivers/dma/xilinx/zynqmp_dma.c                    |  6 +-
 drivers/gpu/drm/drm_pci.c                          |  5 +-
 drivers/infiniband/hw/bnxt_re/qplib_rcfw.c         |  4 +-
 drivers/infiniband/hw/bnxt_re/qplib_res.c          |  8 +--
 drivers/infiniband/hw/cxgb3/cxio_hal.c             |  6 +-
 drivers/infiniband/hw/cxgb4/qp.c                   |  5 +-
 drivers/infiniband/hw/hfi1/init.c                  | 29 ++++-----
 drivers/infiniband/hw/hfi1/pio.c                   |  9 ++-
 drivers/infiniband/hw/hfi1/sdma.c                  | 27 +++------
 drivers/infiniband/hw/hns/hns_roce_alloc.c         | 11 ++--
 drivers/infiniband/hw/hns/hns_roce_hw_v2.c         | 10 ++--
 drivers/infiniband/hw/i40iw/i40iw_utils.c          |  4 +-
 drivers/infiniband/hw/mthca/mthca_memfree.c        |  5 +-
 drivers/infiniband/hw/ocrdma/ocrdma_hw.c           | 14 ++---
 drivers/infiniband/hw/ocrdma/ocrdma_stats.c        |  4 +-
 drivers/infiniband/hw/ocrdma/ocrdma_verbs.c        |  6 +-
 drivers/infiniband/hw/qedr/verbs.c                 |  4 +-
 drivers/infiniband/hw/vmw_pvrdma/pvrdma_main.c     |  4 +-
 drivers/input/touchscreen/raspberrypi-ts.c         |  4 +-
 drivers/iommu/mtk_iommu_v1.c                       |  5 +-
 drivers/media/pci/intel/ipu3/ipu3-cio2.c           |  4 +-
 .../media/platform/mtk-vcodec/mtk_vcodec_util.c    |  2 +-
 drivers/misc/genwqe/card_utils.c                   |  4 +-
 drivers/mmc/host/sdhci.c                           |  5 +-
 drivers/net/ethernet/aeroflex/greth.c              | 12 ++--
 drivers/net/ethernet/alacritech/slicoss.c          | 12 ++--
 drivers/net/ethernet/amazon/ena/ena_com.c          | 61 +++++++++----------
 drivers/net/ethernet/apm/xgene-v2/main.c           |  8 +--
 drivers/net/ethernet/atheros/alx/main.c            |  7 +--
 drivers/net/ethernet/atheros/atl1c/atl1c_main.c    |  4 +-
 drivers/net/ethernet/broadcom/bcm63xx_enet.c       |  8 +--
 drivers/net/ethernet/broadcom/bcmsysport.c         |  4 +-
 drivers/net/ethernet/broadcom/bgmac.c              | 12 ++--
 drivers/net/ethernet/broadcom/bnx2.c               |  4 +-
 drivers/net/ethernet/broadcom/bnx2x/bnx2x.h        |  2 +-
 drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.h    |  2 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt.c          | 16 ++---
 drivers/net/ethernet/broadcom/bnxt/bnxt_dcb.c      |  4 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt_devlink.c  |  4 +-
 drivers/net/ethernet/broadcom/tg3.c                | 22 +++----
 drivers/net/ethernet/cavium/thunder/nicvf_queues.c |  2 +-
 drivers/net/ethernet/chelsio/cxgb3/sge.c           |  2 +-
 drivers/net/ethernet/chelsio/cxgb4/sge.c           |  2 +-
 drivers/net/ethernet/chelsio/cxgb4vf/sge.c         |  2 +-
 drivers/net/ethernet/emulex/benet/be_cmds.c        | 68 +++++++++++-----------
 drivers/net/ethernet/emulex/benet/be_ethtool.c     | 18 +++---
 drivers/net/ethernet/emulex/benet/be_main.c        | 18 +++---
 drivers/net/ethernet/faraday/ftgmac100.c           | 14 ++---
 drivers/net/ethernet/faraday/ftmac100.c            |  7 +--
 drivers/net/ethernet/hisilicon/hix5hd2_gmac.c      |  4 +-
 drivers/net/ethernet/hisilicon/hns3/hns3_enet.c    |  5 +-
 .../net/ethernet/hisilicon/hns3/hns3pf/hclge_cmd.c |  5 +-
 .../ethernet/hisilicon/hns3/hns3vf/hclgevf_cmd.c   |  5 +-
 .../net/ethernet/huawei/hinic/hinic_hw_api_cmd.c   | 16 ++---
 drivers/net/ethernet/huawei/hinic/hinic_hw_eqs.c   |  8 +--
 drivers/net/ethernet/huawei/hinic/hinic_hw_io.c    |  6 +-
 drivers/net/ethernet/huawei/hinic/hinic_hw_qp.c    | 10 ++--
 drivers/net/ethernet/huawei/hinic/hinic_hw_wq.c    |  8 +--
 drivers/net/ethernet/ibm/emac/mal.c                |  4 +-
 drivers/net/ethernet/intel/e1000/e1000_ethtool.c   |  8 +--
 drivers/net/ethernet/intel/e1000e/netdev.c         |  4 +-
 drivers/net/ethernet/intel/i40e/i40e_main.c        |  4 +-
 drivers/net/ethernet/intel/ixgb/ixgb_main.c        |  8 +--
 drivers/net/ethernet/marvell/mvpp2/mvpp2_main.c    |  6 +-
 drivers/net/ethernet/marvell/octeontx2/af/common.h |  2 +-
 drivers/net/ethernet/marvell/pxa168_eth.c          | 18 +++---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c        | 18 +++---
 drivers/net/ethernet/mellanox/mlx4/alloc.c         |  8 +--
 drivers/net/ethernet/mellanox/mlx5/core/alloc.c    |  4 +-
 drivers/net/ethernet/mellanox/mlx5/core/cmd.c      | 10 ++--
 drivers/net/ethernet/myricom/myri10ge/myri10ge.c   |  6 +-
 .../net/ethernet/netronome/nfp/nfp_net_common.c    | 12 ++--
 drivers/net/ethernet/ni/nixge.c                    | 12 ++--
 .../net/ethernet/oki-semi/pch_gbe/pch_gbe_main.c   | 12 ++--
 drivers/net/ethernet/pasemi/pasemi_mac.c           |  6 +-
 drivers/net/ethernet/qlogic/qed/qed_cxt.c          | 16 ++---
 drivers/net/ethernet/qlogic/qlcnic/qlcnic_ctx.c    | 28 ++++-----
 drivers/net/ethernet/qualcomm/emac/emac-mac.c      |  2 +-
 drivers/net/ethernet/samsung/sxgbe/sxgbe_main.c    | 12 ++--
 drivers/net/ethernet/sfc/falcon/nic.c              |  4 +-
 drivers/net/ethernet/sfc/nic.c                     |  4 +-
 drivers/net/ethernet/sgi/meth.c                    |  4 +-
 drivers/net/ethernet/socionext/netsec.c            |  4 +-
 drivers/net/ethernet/stmicro/stmmac/stmmac_main.c  | 40 +++++--------
 drivers/net/ethernet/tundra/tsi108_eth.c           |  8 +--
 drivers/net/ethernet/xilinx/ll_temac_main.c        | 12 ++--
 drivers/net/ethernet/xilinx/xilinx_axienet_main.c  | 12 ++--
 drivers/net/fddi/defxx.c                           |  6 +-
 drivers/net/fddi/skfp/skfddi.c                     |  8 +--
 drivers/net/vmxnet3/vmxnet3_drv.c                  |  8 +--
 drivers/net/wan/fsl_ucc_hdlc.c                     |  7 +--
 drivers/net/wireless/ath/ath10k/ce.c               |  7 +--
 drivers/net/wireless/ath/ath10k/mac.c              |  8 +--
 drivers/net/wireless/ath/ath10k/pci.c              |  3 +-
 drivers/net/wireless/ath/ath10k/wmi.c              |  2 +-
 drivers/net/wireless/ath/wcn36xx/dxe.c             | 17 +++---
 drivers/net/wireless/ath/wil6210/txrx_edma.c       |  8 +--
 drivers/net/wireless/broadcom/b43/dma.c            |  6 +-
 drivers/net/wireless/broadcom/b43legacy/dma.c      |  6 +-
 .../wireless/broadcom/brcm80211/brcmfmac/pcie.c    | 16 ++---
 drivers/net/wireless/intel/iwlwifi/pcie/rx.c       | 39 +++++--------
 drivers/net/wireless/ralink/rt2x00/rt2x00mmio.c    |  6 +-
 drivers/ntb/hw/mscc/ntb_hw_switchtec.c             |  8 +--
 drivers/nvme/host/pci.c                            |  8 +--
 drivers/pci/controller/pcie-iproc-msi.c            |  6 +-
 drivers/pci/switch/switchtec.c                     |  8 +--
 drivers/rapidio/devices/tsi721.c                   | 22 +++----
 drivers/rapidio/devices/tsi721_dma.c               |  8 +--
 drivers/s390/net/ism_drv.c                         | 15 +++--
 drivers/scsi/3w-sas.c                              |  5 +-
 drivers/scsi/a100u2w.c                             |  8 +--
 drivers/scsi/arcmsr/arcmsr_hba.c                   | 18 ++++--
 drivers/scsi/be2iscsi/be_main.c                    |  4 +-
 drivers/scsi/be2iscsi/be_mgmt.c                    | 11 ++--
 drivers/scsi/bfa/bfad_bsg.c                        |  6 +-
 drivers/scsi/bnx2fc/bnx2fc_hwi.c                   | 49 ++++++++--------
 drivers/scsi/bnx2fc/bnx2fc_tgt.c                   | 44 +++++++-------
 drivers/scsi/bnx2i/bnx2i_hwi.c                     |  8 +--
 drivers/scsi/csiostor/csio_wr.c                    |  4 +-
 drivers/scsi/lpfc/lpfc_bsg.c                       |  4 +-
 drivers/scsi/lpfc/lpfc_init.c                      | 14 ++---
 drivers/scsi/lpfc/lpfc_mbox.c                      |  6 +-
 drivers/scsi/lpfc/lpfc_sli.c                       | 15 +++--
 drivers/scsi/megaraid/megaraid_mbox.c              | 15 ++---
 drivers/scsi/megaraid/megaraid_sas_base.c          | 26 ++++-----
 drivers/scsi/megaraid/megaraid_sas_fusion.c        |  5 +-
 drivers/scsi/mesh.c                                |  5 +-
 drivers/scsi/mvumi.c                               |  9 +--
 drivers/scsi/pm8001/pm8001_sas.c                   |  4 +-
 drivers/scsi/qedf/qedf_main.c                      | 29 +++++----
 drivers/scsi/qedi/qedi_main.c                      | 39 ++++++-------
 drivers/scsi/qla2xxx/qla_attr.c                    |  4 +-
 drivers/scsi/qla2xxx/qla_bsg.c                     |  4 +-
 drivers/scsi/qla2xxx/qla_gs.c                      | 14 +++--
 drivers/scsi/qla2xxx/qla_init.c                    |  8 +--
 drivers/scsi/qla4xxx/ql4_init.c                    |  4 +-
 drivers/scsi/qla4xxx/ql4_mbx.c                     | 18 +++---
 drivers/scsi/qla4xxx/ql4_nx.c                      |  4 +-
 drivers/scsi/qla4xxx/ql4_os.c                      | 10 ++--
 drivers/scsi/smartpqi/smartpqi_init.c              | 32 +++++-----
 drivers/soc/fsl/qbman/dpaa_sys.c                   |  2 +-
 drivers/spi/spi-pic32-sqi.c                        |  6 +-
 drivers/staging/mt7621-eth/mtk_eth_soc.c           |  3 +-
 .../interface/vchiq_arm/vchiq_2835_arm.c           |  6 +-
 drivers/staging/vt6655/device_main.c               | 19 ++----
 drivers/usb/gadget/udc/bdc/bdc_core.c              | 13 ++---
 drivers/usb/host/uhci-hcd.c                        |  6 +-
 drivers/usb/host/xhci-mem.c                        |  8 +--
 drivers/video/fbdev/da8xx-fb.c                     |  6 +-
 include/linux/dma-mapping.h                        |  9 ---
 include/linux/pci-dma-compat.h                     |  2 +-
 scripts/coccinelle/api/alloc/alloc_cast.cocci      |  8 +--
 scripts/coccinelle/api/alloc/zalloc-simple.cocci   | 11 +---
 sound/aoa/soundbus/i2sbus/core.c                   |  4 +-
 sound/sparc/dbri.c                                 |  4 +-
 179 files changed, 923 insertions(+), 975 deletions(-)
#1 lore raw
Quoted 3 lines · Christoph Hellwig

Lets see if this works better as a pull request than the plain patches:

So I'm still debating this myself, which is why I haven't reacted.

The reason I'm not entirely sure this makes sense is that not every user actually mmaps things into user space.

And now if somebody for some reason doesn't want the overhead, there is really *no* way to say that.

Maybe nobody cares. But from what I can tell, the mmap case is actually really quite rare (well, at least the regular "dma_mmap_attrs/coherent()" kind).

I'd be happier if there at least was an interface to specify "yes, I want zeroing". But there isn't. You did not, for example, just make __GFP_ZERO work. No. You just said "GFP_ZERO hasn't worked reliably, so we had that other interface, and now we're doing it unconditionally, so we can remove the other interface".

This all *despite* the fact that as far as I can tell, 95% of the users don't actually need or care about zeroing and don't really want it.

So explain that.

Linus

#2 lore raw
Quoted 4 lines · Linus Torvalds

So I'm still debating this myself, which is why I haven't reacted.

The reason I'm not entirely sure this makes sense is that not every user actually mmaps things into user space.

While that is the security reason for it, we also had quite a few drivers just plain relying on the fact and breaking on architectures where it didn't work.

Quoted 9 lines

I'd be happier if there at least was an interface to specify "yes, I want zeroing". But there isn't. You did not, for example, just make __GFP_ZERO work. No. You just said "GFP_ZERO hasn't worked reliably, so we had that other interface, and now we're doing it unconditionally, so we can remove the other interface".

This all *despite* the fact that as far as I can tell, 95% of the users don't actually need or care about zeroing and don't really want it.

Many users don't need it for security reasons, but given that x86 and arm have dont it forever various drivers started relying on the behavior. That combined with the fact that DMA coherent allocation are a slow path interface it really doesn't make much sense to micro-optimize it and potentially break users. If we have a user that really cares about not zeroing we can just offer a DMA_ATTR_SKIP_ZEROING flag for dma_alloc_attrs and switch the user that that cares to it. But I don't really think it is needed.

#3 lore raw
Quoted 4 lines · Christoph Hellwig

Many users don't need it for security reasons, but given that x86 and arm have done it forever various drivers started relying on the behavior.

Ok, I guess that's a pretty strong argument.

Linus

merge bot Merged
#4 lore raw
In reply to Christoph Hellwig
The pull request you sent on Sat, 12 Jan 2019 19:13:55 +0100:

The pull request you sent on Sat, 12 Jan 2019 19:13:55 +0100:

Quoted 1 line

git://git.infradead.org/users/hch/dma-mapping.git tags/remove-dma_zalloc_coherent-5.0

has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/66c56cfa64d9dbb9efa8a06c1aece77e8d57ea19

Thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.wiki.kernel.org/userdoc/prtracker