[<prev] [next>] [day] [month] [year] [list]
Message-ID: <202207221447.1b5HzBgh-lkp@intel.com>
Date: Fri, 22 Jul 2022 14:47:45 +0800
From: kernel test robot <lkp@...el.com>
To: Keith Busch <kbusch@...nel.org>
Cc: llvm@...ts.linux.dev, kbuild-all@...ts.01.org,
linux-kernel@...r.kernel.org
Subject: [kbusch:dma-prereg-v3-wip 6/6] drivers/nvme/host/pci.c:967:7:
warning: variable 'nprps_left' is uninitialized when used here
tree: https://git.kernel.org/pub/scm/linux/kernel/git/kbusch/linux.git dma-prereg-v3-wip
head: 4625fa7adf745334c0a2eb635121d587df797b29
commit: 4625fa7adf745334c0a2eb635121d587df797b29 [6/6] nvme-pci: implement dma_map support
config: x86_64-randconfig-a014 (https://download.01.org/0day-ci/archive/20220722/202207221447.1b5HzBgh-lkp@intel.com/config)
compiler: clang version 15.0.0 (https://github.com/llvm/llvm-project 72686d68c137551cce816416190a18d45b4d4e2a)
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# https://git.kernel.org/pub/scm/linux/kernel/git/kbusch/linux.git/commit/?id=4625fa7adf745334c0a2eb635121d587df797b29
git remote add kbusch https://git.kernel.org/pub/scm/linux/kernel/git/kbusch/linux.git
git fetch --no-tags kbusch dma-prereg-v3-wip
git checkout 4625fa7adf745334c0a2eb635121d587df797b29
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash drivers/nvme/host/
If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <lkp@...el.com>
All warnings (new ones prefixed by >>):
>> drivers/nvme/host/pci.c:967:7: warning: variable 'nprps_left' is uninitialized when used here [-Wuninitialized]
if (nprps_left <= last_prp + 1) {
^~~~~~~~~~
drivers/nvme/host/pci.c:893:49: note: initialize the variable 'nprps_left' to silence this warning
int index, offset, i, length, nprps, nprps_left;
^
= 0
drivers/nvme/host/pci.c:1935:7: warning: variable 'pages_per' set but not used [-Wunused-but-set-variable]
int pages_per = nvme_pages;
^
2 warnings generated.
vim +/nprps_left +967 drivers/nvme/host/pci.c
886
887 static blk_status_t nvme_premapped(struct nvme_dev *dev, struct request *req,
888 struct nvme_rw_command *cmnd,
889 struct nvme_iod *iod)
890 {
891 static const int last_prp = NVME_CTRL_PAGE_SIZE / sizeof(__le64) - 1;
892 dma_addr_t prp_list_start, prp_list_end, prp_dma;
893 int index, offset, i, length, nprps, nprps_left;
894 void **list = nvme_pci_iod_list(req);
895 struct nvme_dma_mapping *mapping;
896 struct dma_pool *pool;
897 __le64 *prp_list;
898 bool needs_sync;
899
900 mapping = blk_rq_dma_tag(req);
901 offset = blk_rq_dma_offset(req) + mapping->offset;
902 index = offset >> NVME_CTRL_PAGE_SHIFT;
903 offset = offset & (NVME_CTRL_PAGE_SIZE - 1);
904 needs_sync = rq_data_dir(req) == WRITE &&
905 dma_need_sync(dev->dev, le64_to_cpu(mapping->prps[index]));
906
907 /*
908 * XXX: For PAGE_SIZE > NVME_CTRL_PAGE_SIZE, is it faster to save the
909 * PRP list implementation and sync multiple partial pages, more
910 * efficient to sync PAGE_SIZE and build the PRP list per-IO from a
911 * host PAGE_SIZE representation, or cleverly sync physically
912 * contiguous regions?
913 */
914 if (needs_sync) {
915 dma_sync_single_for_device(dev->dev,
916 le64_to_cpu(mapping->prps[index]),
917 NVME_CTRL_PAGE_SIZE - offset, DMA_TO_DEVICE);
918 }
919
920 length = blk_rq_payload_bytes(req) - (NVME_CTRL_PAGE_SIZE - offset);
921 cmnd->dptr.prp1 = cpu_to_le64(le64_to_cpu(mapping->prps[index++]) + offset);
922
923 if (length <= 0)
924 return BLK_STS_OK;
925
926 if (length <= NVME_CTRL_PAGE_SIZE) {
927 if (needs_sync)
928 dma_sync_single_for_device(dev->dev,
929 le64_to_cpu(mapping->prps[index]),
930 NVME_CTRL_PAGE_SIZE, DMA_TO_DEVICE);
931 cmnd->dptr.prp2 = mapping->prps[index];
932 return BLK_STS_OK;
933 }
934
935 nprps = DIV_ROUND_UP(length, NVME_CTRL_PAGE_SIZE);
936 prp_list_start = mapping->prp_dma_addr + 8 * index;
937 prp_list_end = prp_list_start + 8 * nprps;
938
939 /* Optimization when remaining list fits in one nvme page */
940 if ((prp_list_start >> NVME_CTRL_PAGE_SHIFT) ==
941 (prp_list_end >> NVME_CTRL_PAGE_SHIFT)) {
942 cmnd->dptr.prp2 = cpu_to_le64(prp_list_start);
943 goto sync;
944 }
945
946 if (nprps <= (256 / 8)) {
947 pool = dev->prp_small_pool;
948 iod->npages = 0;
949 } else {
950 pool = dev->prp_page_pool;
951 iod->npages = 1;
952 }
953
954 prp_list = dma_pool_alloc(pool, GFP_ATOMIC, &prp_dma);
955 if (!prp_list) {
956 iod->npages = -1;
957 return BLK_STS_RESOURCE;
958 }
959
960 list[0] = prp_list;
961 iod->first_dma = prp_dma;
962 i = 0;
963 for (;;) {
964 dma_addr_t next_prp_dma;
965 __le64 *next_prp_list;
966
> 967 if (nprps_left <= last_prp + 1) {
968 memcpy(prp_list, &mapping->prps[index], nprps_left * 8);
969 break;
970 }
971
972 memcpy(prp_list, &mapping->prps[index],
973 NVME_CTRL_PAGE_SIZE - 8);
974 nprps_left -= last_prp;
975 index += last_prp;
976
977 next_prp_list = dma_pool_alloc(pool, GFP_ATOMIC, &next_prp_dma);
978 if (!next_prp_list)
979 goto free_prps;
980
981 prp_list[last_prp] = cpu_to_le64(next_prp_dma);
982 prp_list = next_prp_list;
983 prp_dma = next_prp_dma;
984 list[iod->npages++] = prp_list;
985 }
986 cmnd->dptr.prp2 = cpu_to_le64(iod->first_dma);
987
988 sync:
989 if (!needs_sync)
990 return BLK_STS_OK;
991
992 for (i = 0; i < nprps; i++)
993 dma_sync_single_for_device(dev->dev,
994 le64_to_cpu(mapping->prps[index++]),
995 NVME_CTRL_PAGE_SIZE, DMA_TO_DEVICE);
996 return BLK_STS_OK;
997
998 free_prps:
999 nvme_free_prps(dev, req);
1000 return BLK_STS_RESOURCE;
1001 }
1002
--
0-DAY CI Kernel Test Service
https://01.org/lkp
Powered by blists - more mailing lists