| Message ID | 20230329204652.52785-1-kch@nvidia.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 3/29/23 2:46 PM, Chaitanya Kulkarni wrote: > Hi, > > From :include/linux/highmem.h: > "kmap_atomic - Atomically map a page for temporary usage - Deprecated!" > > Use memcpy_from_page() since does the same job of mapping, copying, and > unmaping except it uses non deprecated kmap_local_page() and > kunmap_local(). Following are the differences between kmal_local_page() > and kmap_atomic() :- Looks fine to me, but I'd fold patches 1-3 rather than split them up.
On 3/29/23 15:48, Jens Axboe wrote: > On 3/29/23 2:46 PM, Chaitanya Kulkarni wrote: >> Hi, >> >> From :include/linux/highmem.h: >> "kmap_atomic - Atomically map a page for temporary usage - Deprecated!" >> >> Use memcpy_from_page() since does the same job of mapping, copying, and >> unmaping except it uses non deprecated kmap_local_page() and >> kunmap_local(). Following are the differences between kmal_local_page() >> and kmap_atomic() :- > Looks fine to me, but I'd fold patches 1-3 rather than split them up. > Sent V2 with above comment, first three patches are from different code path and they are doing unrelated changes:- 1. WRITE :- copy_to_nullb() only use memcpy_page(). 2. READ :- copy_from_nullb() only use memcapy_page() and zero_user(). 3. I guess zoned read beyond write pointer null_fill_pattern() memset_page(). if anything goes wrong in any of 3 code paths we will have to entire change which we shouldn't, that's why kept it separate, I'm fine with whatever you decide ... -ck
On 3/30/23 12:56?PM, Chaitanya Kulkarni wrote: > On 3/29/23 15:48, Jens Axboe wrote: >> On 3/29/23 2:46?PM, Chaitanya Kulkarni wrote: >>> Hi, >>> >>> From :include/linux/highmem.h: >>> "kmap_atomic - Atomically map a page for temporary usage - Deprecated!" >>> >>> Use memcpy_from_page() since does the same job of mapping, copying, and >>> unmaping except it uses non deprecated kmap_local_page() and >>> kunmap_local(). Following are the differences between kmal_local_page() >>> and kmap_atomic() :- >> Looks fine to me, but I'd fold patches 1-3 rather than split them up. >> > > Sent V2 with above comment, first three patches are from > different code path and they are doing unrelated changes:- > > 1. WRITE :- copy_to_nullb() only use memcpy_page(). > 2. READ :- copy_from_nullb() only use memcapy_page() and zero_user(). > 3. I guess zoned read beyond write pointer null_fill_pattern() > memset_page(). > > if anything goes wrong in any of 3 code paths we will have to > entire change which we shouldn't, that's why kept it separate, > I'm fine with whatever you decide ... It doesn't matter... It's not like this is a hugely complicated thing. It's just a straight forward conversion. They all just switch to using a helper, which is the right thing to do as it reduces the stuff we have to do in there. Only reason why I excluded patch 4 is that it is a bit different than the others. But it really could just be one patch.
diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c index bc2c58724df3..46b3381cd42b 100644 --- a/drivers/block/null_blk/main.c +++ b/drivers/block/null_blk/main.c @@ -1030,6 +1030,7 @@ static int null_flush_cache_page(struct nullb *nullb, struct nullb_page *c_page) if (!t_page) return -ENOMEM; + pr_info("%s %d kmap_local_page()\n", __func__, __LINE__); src = kmap_local_page(c_page->page); dst = kmap_local_page(t_page->page); @@ -1125,6 +1126,7 @@ static int copy_to_nullb(struct nullb *nullb, struct page *source, if (!t_page) return -ENOSPC; + pr_info("%s %d memcpy_page()\n", __func__, __LINE__); memcpy_page(t_page->page, offset, source, off + count, temp); __set_bit(sector & SECTOR_MASK, t_page->bitmap); @@ -1152,11 +1154,14 @@ static int copy_from_nullb(struct nullb *nullb, struct page *dest, t_page = null_lookup_page(nullb, sector, false, !null_cache_active(nullb)); - if (t_page) + if (t_page) { memcpy_page(dest, off + count, t_page->page, offset, temp); - else + pr_info("%s %d memcpy_page()\n", __func__, __LINE__); + } else { zero_user(dest, off + count, temp); + pr_info("%s %d zero_user()\n", __func__, __LINE__); + } count += temp; sector += temp >> SECTOR_SHIFT; @@ -1167,6 +1172,7 @@ static int copy_from_nullb(struct nullb *nullb, struct page *dest, static void nullb_fill_pattern(struct nullb *nullb, struct page *page, unsigned int len, unsigned int off) { + pr_info("%s %d memset_page()\n", __func__, __LINE__); memset_page(page, off, 0xff, len); }
Hi, From :include/linux/highmem.h: "kmap_atomic - Atomically map a page for temporary usage - Deprecated!" Use memcpy_from_page() since does the same job of mapping, copying, and unmaping except it uses non deprecated kmap_local_page() and kunmap_local(). Following are the differences between kmal_local_page() and kmap_atomic() :- * creates local mapping per thread, local to CPU & not globally visible * allows to be called from any context * allows task preemption There is a slight performance difference observed with the use of new API on the one arch I've tested with two different sets :- Set 1 (Average of 3 runs) :- ----------------------------- * Latency (lower is better) :- approximately the same * IOPS/BW (higner is better) :- ~12k higner with this patch series * CPU Usage (lower is better) :- approximately the same Set 2 (Average of 3 runs) :- ----------------------------- * Latency (lower is better) :- ~116 lower with this patch seires * IOPS/BW (higner is better) :- ~5k higner with this patch series * CPU Usage (lower is better) :- approximately the same Below is the test for the fio verification job and perf numbers on nullb along with the test log with debug patch to cover all the calls that are modified in this series:- copy_to_nullb() memcpy_to_page() copy_from_nullb() memcpy_to_page() zero_user() null_fill_pattern() memset_page() null_flush_cache_page() kmap_local_page() kunmap_local_page() In case someone shows up with performance regression on the arch that I've don't have access to we can decide then if we want to drop it this series or keep using deprecated kernel API, but I think removing deprecated API is useful in long term in anyway. -ck Chaitanya Kulkarni (4): null_blk: use memcpy_page() in copy_to_nullb() null_blk: use memcpy_page() in copy_from_nullb() null-blk: use memset_page() to fill page pattern null_blk: use kmap_local_page() and kunmap_local() drivers/block/null_blk/main.c | 37 +++++++++++------------------------ 1 file changed, 11 insertions(+), 26 deletions(-) ####################################################################### Performance numbers :- * Set 1:- ---------------- * Avg Latency delta (lower is better) :- ~14 higher with this patch seires linux-block (nullb-memcpy) # grep -w "lat (usec):" *null*fio default-nullb.1.fio: lat (usec): min=1794, max=2179.1k, avg=23.26, stdev=5605.88 default-nullb.2.fio: lat (usec): min=1995, max=6038.1k, avg=24.21, stdev=6305.29 default-nullb.3.fio: lat (usec): min=2, max=5554, avg=24.12, stdev= 6.27 (23.26+24.21+24.12)/3 = 23 with-memcpy-nullb.1.fio: lat (usec): min=2, max=2461, avg=23.49, stdev= 5.30 with-memcpy-nullb.2.fio: lat (usec): min=1943, max=2500.7k, avg=23.53, stdev=5942.17 with-memcpy-nullb.3.fio: lat (usec): min=2, max=2316, avg=23.64, stdev= 5.58 (23.49+23.53+23.64)/3 = 23 * Ave IOPS/BW delta (higner is better):- ~47k higner with this patch series linux-block (nullb-memcpy) # grep IOPS *nullb*fio default-nullb.1.fio: read: IOPS=1017k, BW=3974MiB/s (4167MB/s)(233GiB/60002msec) default-nullb.2.fio: read: IOPS=978k, BW=3820MiB/s (4006MB/s)(224GiB/60001msec) default-nullb.3.fio: read: IOPS=982k, BW=3835MiB/s (4022MB/s)(225GiB/60002msec) (1017+978+982)/3 = 992k with-memcpy-nullb.1.fio: read: IOPS=1008k, BW=3939MiB/s (4130MB/s)(231GiB/60002msec) with-memcpy-nullb.2.fio: read: IOPS=1006k, BW=3931MiB/s (4122MB/s)(230GiB/60001msec) with-memcpy-nullb.3.fio: read: IOPS=1002k, BW=3913MiB/s (4103MB/s)(229GiB/60001msec) (1008+1006+1002)/3 = 1005; * Avg CPU Usage delta (lower is better) :- approximately the same linux-block (nullb-memcpy) # grep cpu *nullb*fio default-nullb.1.fio: cpu: usr=2.08%, sys=97.83%, ctx=4767, majf=0, minf=367 default-nullb.2.fio: cpu: usr=2.06%, sys=97.84%, ctx=3013, majf=0, minf=379 default-nullb.3.fio: cpu: usr=2.15%, sys=97.75%, ctx=17878, majf=0, minf=369 (97.83+97.84+97.75)/3 = 97 with-memcpy-nullb.1.fio: cpu: usr=2.08%, sys=97.83%, ctx=8507, majf=0, minf=391 with-memcpy-nullb.2.fio: cpu: usr=2.12%, sys=97.79%, ctx=4817, majf=0, minf=359 with-memcpy-nullb.3.fio: cpu: usr=2.20%, sys=97.72%, ctx=2963, majf=0, minf=351 (97.83+97.79+97.72)/3 = 97 * Set 2:- ---------------- * Avg Latency delta (lower is better) :- ~14 higher with this patch seires linux-block (nullb-memcpy) # grep -w "lat (nsec):" *null*fio default-nullb.1.fio: lat (nsec): min=1975, max=4169.3k, avg=22580.36, stdev=4611.23 default-nullb.2.fio: lat (nsec): min=1984, max=1055.5k, avg=22609.24, stdev=4611.20 default-nullb.3.fio: lat (nsec): min=1974, max=2565.9k, avg=22750.96, stdev=4480.30 (22580.36+22609.24+22750.96)/3 = 22646 with-memcpy-nullb.1.fio: lat (nsec): min=1983, max=11398k, avg=22441.78, stdev=6809.98 with-memcpy-nullb.2.fio: lat (nsec): min=1853, max=1083.9k, avg=22620.39, stdev=4607.59 with-memcpy-nullb.3.fio: lat (nsec): min=1974, max=6048.1k, avg=22528.70, stdev=4902.61 (22441.78+22620.39+22528.70)/3 = 22530 * Ave IOPS/BW delta (higner is better ):- ~47k higner with this patch series linux-block (nullb-memcpy) # grep IOPS *null*fio default-nullb.1.fio: read: IOPS=1050k, BW=4100MiB/s (4299MB/s)(240GiB/60002msec) default-nullb.2.fio: read: IOPS=1048k, BW=4095MiB/s (4294MB/s)(240GiB/60001msec) default-nullb.3.fio: read: IOPS=1042k, BW=4069MiB/s (4267MB/s)(238GiB/60001msec) (1050+1048+1042)/3 = 1046 with-memcpy-nullb.1.fio: read: IOPS=1055k, BW=4123MiB/s (4323MB/s)(242GiB/60001msec) with-memcpy-nullb.2.fio: read: IOPS=1048k, BW=4093MiB/s (4292MB/s)(240GiB/60001msec) with-memcpy-nullb.3.fio: read: IOPS=1051k, BW=4107MiB/s (4306MB/s)(241GiB/60001msec) (1055+1048+1051)/3 = 1051 * Avg CPU Usage delta (lower is better) :- approximately the same linux-block (nullb-memcpy) # grep cpu *nullb*fio default-nullb.1.fio: cpu: usr=1.99%, sys=97.94%, ctx=4581, majf=0, minf=353 default-nullb.2.fio: cpu: usr=1.99%, sys=97.94%, ctx=7169, majf=0, minf=373 default-nullb.3.fio: cpu: usr=2.02%, sys=97.91%, ctx=3560, majf=0, minf=348 (97.94+97.94+97.91)/3 = 97 with-memcpy-nullb.1.fio: cpu: usr=2.03%, sys=97.91%, ctx=2405, majf=0, minf=348 with-memcpy-nullb.2.fio: cpu: usr=2.01%, sys=97.92%, ctx=2313, majf=0, minf=364 with-memcpy-nullb.3.fio: cpu: usr=2.04%, sys=97.88%, ctx=7224, majf=0, minf=348 (97.91+97.92+97.88)/3 = 97 ####################################################################### Testing with fio verification and randread workload on null:- linux-block (nullb-memcpy) # sh test-nullb-memcpy-perf.sh Already on 'nullb-memcpy' 0eb5ac9f7c03 (HEAD -> nullb-memcpy, brd-memcpy) null_blk: use kmap_local_page() and kunmap_local() 9a8f884aea9e null-blk: use memset_page() to fill page pattern 5363fb1a0481 null_blk: use memcpy_page() in copy_from_nullb() 66dee4a65d8c null_blk: use memcpy_page() in copy_to_nullb() + umount /mnt/nullb0 umount: /mnt/nullb0: not mounted. + rmdir config/nullb/nullb0 + dmesg -c + modprobe -r null_blk + lsmod + grep null_blk ++ nproc + make -j 48 M=drivers/block modules CC [M] drivers/block/null_blk/main.o LD [M] drivers/block/null_blk/null_blk.o MODPOST drivers/block/Module.symvers LD [M] drivers/block/null_blk/null_blk.ko + HOST=drivers/block/null_blk/ ++ uname -r + HOST_DEST=/lib/modules/6.3.0-rc4lblk+/kernel/drivers/block/null_blk/ + cp drivers/block/null_blk//null_blk.ko /lib/modules/6.3.0-rc4lblk+/kernel/drivers/block/null_blk// + ls -lrth /lib/modules/6.3.0-rc4lblk+/kernel/drivers/block/null_blk//null_blk.ko -rw-r--r--. 1 root root 1.2M Mar 28 22:19 /lib/modules/6.3.0-rc4lblk+/kernel/drivers/block/null_blk//null_blk.ko + sleep 1 + dmesg -c + let NN=1-1 + modprobe -r null_blk + modprobe null_blk nr_devices=0 + echo loading devices loading devices ++ seq 0 0 + for i in `seq 0 $NN` + NULLB_DIR=config/nullb/nullb0 + mkdir config/nullb/nullb0 + cat config/nullb/nullb0/zoned 0 + echo 256 + echo 0 + echo 1 + echo 4096 + echo 1024 + echo 0 + echo 1 + echo 2 + cat config/nullb/nullb0/queue_mode 2 + echo 1 ++ cat config/nullb/nullb0/index + IDX=0 + echo -n ' 0 ' 0 + sleep .50 + lsblk + grep null + sort nullb0 250:0 0 1G 0 disk + sleep 1 + dmesg -c [109428.330925] null_blk: module loaded [109428.339399] null_blk: disk nullb0 created + lsblk + grep null nullb0 250:0 0 1G 0 disk + dmesg -c NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINT sda 8:0 0 50G 0 disk ├─sda1 8:1 0 1G 0 part /boot └─sda2 8:2 0 49G 0 part /home sdb 8:16 0 100G 0 disk /mnt/data sr0 11:0 1 1024M 0 rom nullb0 250:0 0 1G 0 disk zram0 251:0 0 8G 0 disk [SWAP] vda 252:0 0 512M 0 disk nvme0n1 259:0 0 1G 0 disk write-and-verify: (g=0): rw=randwrite, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=16 fio-3.27 Starting 1 process Jobs: 1 (f=1): [V(1)][52.6%][r=551MiB/s,w=451MiB/s][r=141k,w=115k IOPS][eta 00m:18s] write-and-verify: (groupid=0, jobs=1): err= 0: pid=32911: Tue Mar 28 22:23:41 2023 read: IOPS=348k, BW=1358MiB/s (1424MB/s)(1024MiB/754msec) slat (nsec): min=1432, max=51208, avg=1796.19, stdev=435.47 clat (nsec): min=1333, max=94580, avg=43406.05, stdev=2272.20 lat (nsec): min=3147, max=96333, avg=45246.12, stdev=2322.51 clat percentiles (nsec): | 1.00th=[41216], 5.00th=[41728], 10.00th=[42240], 20.00th=[42240], | 30.00th=[42752], 40.00th=[42752], 50.00th=[42752], 60.00th=[43264], | 70.00th=[43264], 80.00th=[43776], 90.00th=[44288], 95.00th=[48896], | 99.00th=[52480], 99.50th=[54528], 99.90th=[58112], 99.95th=[59648], | 99.99th=[86528] write: IOPS=202k, BW=788MiB/s (826MB/s)(15.0GiB/19488msec); 0 zone resets slat (usec): min=2, max=824, avg= 4.35, stdev= 1.01 clat (nsec): min=1192, max=920392, avg=74631.98, stdev=5773.07 lat (usec): min=4, max=924, avg=79.03, stdev= 6.03 clat percentiles (usec): | 1.00th=[ 64], 5.00th=[ 71], 10.00th=[ 72], 20.00th=[ 73], | 30.00th=[ 74], 40.00th=[ 74], 50.00th=[ 75], 60.00th=[ 75], | 70.00th=[ 76], 80.00th=[ 76], 90.00th=[ 80], 95.00th=[ 84], | 99.00th=[ 92], 99.50th=[ 100], 99.90th=[ 131], 99.95th=[ 143], | 99.99th=[ 163] bw ( KiB/s): min=743912, max=911200, per=99.94%, avg=806596.92, stdev=22455.14, samples=39 iops : min=185978, max=227800, avg=201649.23, stdev=5613.87, samples=39 lat (usec) : 2=0.01%, 10=0.01%, 20=0.01%, 50=5.97%, 100=93.56% lat (usec) : 250=0.46%, 500=0.01%, 1000=0.01% cpu : usr=53.05%, sys=46.87%, ctx=33, majf=0, minf=6167 IO depths : 1=0.1%, 2=0.1%, 4=0.1%, 8=0.1%, 16=100.0%, 32=0.0%, >=64=0.0% submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0% complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.1%, 32=0.0%, 64=0.0%, >=64=0.0% issued rwts: total=262144,3932160,0,0 short=0,0,0,0 dropped=0,0,0,0 latency : target=0, window=0, percentile=100.00%, depth=16 Run status group 0 (all jobs): READ: bw=1358MiB/s (1424MB/s), 1358MiB/s-1358MiB/s (1424MB/s-1424MB/s), io=1024MiB (1074MB), run=754-754msec WRITE: bw=788MiB/s (826MB/s), 788MiB/s-788MiB/s (826MB/s-826MB/s), io=15.0GiB (16.1GB), run=19488-19488msec Disk stats (read/write): nullb0: ios=228219/3932160, merge=0/0, ticks=172/4509, in_queue=4682, util=99.56% ####################################################################### Testing for each patch with debug patch to cover all modified calls in this patch series :- * null_flush_cache_page() write to null_blk with cache enabled. linux-block (nullb-memcpy) # modprobe null_blk gb=1 cache_size=1 memory_backed=1 linux-block (nullb-memcpy) # dd if=/dev/zero of=/dev/nullb0 count=$((2*1024*1024/4096)) bs=4k 512+0 records in 512+0 records out 2097152 bytes (2.1 MB, 2.0 MiB) copied, 0.00539398 s, 389 MB/s linux-block (nullb-memcpy) # dmesg -c [115885.171915] null_blk: disk nullb0 created [115885.171919] null_blk: module loaded [115889.623978] null_blk: null_flush_cache_page 1033 kmap_local_page() [115889.623988] null_blk: null_flush_cache_page 1033 kmap_local_page() [115889.623991] null_blk: null_flush_cache_page 1033 kmap_local_page() [115889.623993] null_blk: null_flush_cache_page 1033 kmap_local_page() [115889.623996] null_blk: null_flush_cache_page 1033 kmap_local_page() [115889.623998] null_blk: null_flush_cache_page 1033 kmap_local_page() -----------------------------------cut-------------------------------- * null_blk_fill_pattern() read beyond the zone write pointer :- linux-block (nullb-memcpy) # modprobe null_blk gb=1 memory_backed=1 zoned=1 linux-block (nullb-memcpy) # dd if=/dev/nullb0 of=/dev/null bs=4k count=1 1+0 records in 1+0 records out 4096 bytes (4.1 kB, 4.0 KiB) copied, 0.000224737 s, 18.2 MB/s linux-block (nullb-memcpy) # blkzone report /dev/nullb0 start: 0x000000000, len 0x080000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)] start: 0x000080000, len 0x080000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)] start: 0x000100000, len 0x080000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)] start: 0x000180000, len 0x080000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)] linux-block (nullb-memcpy) # dd if=/dev/nullb0 of=/dev/null count=10 bs=4k 10+0 records in 10+0 records out 40960 bytes (41 kB, 40 KiB) copied, 0.000398227 s, 103 MB/s linux-block (nullb-memcpy) # dmesg -c [116117.786823] null_blk: nullb_fill_pattern 1175 memset_page() [116117.786857] null_blk: nullb_fill_pattern 1175 memset_page() [116117.786912] null_blk: disk nullb0 created [116117.786914] null_blk: module loaded [116120.432254] null_blk: nullb_fill_pattern 1175 memset_page() [116120.432259] null_blk: nullb_fill_pattern 1175 memset_page() [116120.432261] null_blk: nullb_fill_pattern 1175 memset_page() [116120.432263] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441410] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441418] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441421] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441423] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441490] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441492] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441494] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441497] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441499] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441500] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441502] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441504] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441509] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441511] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441513] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441515] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441517] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441519] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441521] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441523] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441525] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441527] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441529] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441531] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441534] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441536] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441538] null_blk: nullb_fill_pattern 1175 memset_page() [116129.441539] null_blk: nullb_fill_pattern 1175 memset_page() * copy_to_nullb()/copy from_nullb() Read/write to memory backed null_blk linux-block (nullb-memcpy) # dd if=/dev/zero of=/dev/nullb0 count=1 bs=4k oflag=direct 1+0 records in 1+0 records out 4096 bytes (4.1 kB, 4.0 KiB) copied, 0.000262448 s, 15.6 MB/s linux-block (nullb-memcpy) # dd if=/dev/nullb0 of=/dev/null count=1 bs=4k 1+0 records in 1+0 records out 4096 bytes (4.1 kB, 4.0 KiB) copied, 0.000320249 s, 12.8 MB/s linux-block (nullb-memcpy) # dmesg -c [116147.220358] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220366] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220368] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220369] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220370] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220371] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220372] null_blk: copy_to_nullb 1129 memcpy_page() [116147.220373] null_blk: copy_to_nullb 1129 memcpy_page() [116150.454379] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454388] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454391] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454393] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454395] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454396] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454398] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454400] null_blk: copy_from_nullb 1160 memcpy_page() [116150.454401] null_blk: nullb_fill_pattern 1175 memset_page() [116150.454404] null_blk: nullb_fill_pattern 1175 memset_page() [116150.454407] null_blk: nullb_fill_pattern 1175 memset_page() linux-block (nullb-memcpy) # modprobe -r null_blk * copy_from_nullb() read from membacked null_blk without any writes linux-block (nullb-memcpy) # modprobe null_blk gb=1 memory_backed=1 cache_size=1 linux-block (nullb-memcpy) # dd if=/dev/nullb0 of=/dev/null bs=4k count=1 1+0 records in 1+0 records out 4096 bytes (4.1 kB, 4.0 KiB) copied, 0.000218966 s, 18.7 MB/s linux-block (nullb-memcpy) # dmesg -c [116159.744004] null_blk: copy_from_nullb 1163 zero_user() [116159.744009] null_blk: copy_from_nullb 1163 zero_user() [116159.744009] null_blk: copy_from_nullb 1163 zero_user() [116159.744010] null_blk: copy_from_nullb 1163 zero_user() [116159.744010] null_blk: copy_from_nullb 1163 zero_user() [116159.744011] null_blk: copy_from_nullb 1163 zero_user() [116159.744011] null_blk: copy_from_nullb 1163 zero_user() [116159.744012] null_blk: copy_from_nullb 1163 zero_user() [116159.744021] null_blk: copy_from_nullb 1163 zero_user() [116159.744022] null_blk: copy_from_nullb 1163 zero_user() [116159.744022] null_blk: copy_from_nullb 1163 zero_user() [116159.744023] null_blk: copy_from_nullb 1163 zero_user() [116159.744023] null_blk: copy_from_nullb 1163 zero_user() [116159.744024] null_blk: copy_from_nullb 1163 zero_user() [116159.744024] null_blk: copy_from_nullb 1163 zero_user() [116159.744025] null_blk: copy_from_nullb 1163 zero_user() [116159.744046] null_blk: disk nullb0 created [116159.744047] null_blk: module loaded [116163.094839] null_blk: copy_from_nullb 1163 zero_user() [116163.094847] null_blk: copy_from_nullb 1163 zero_user() [116163.094848] null_blk: copy_from_nullb 1163 zero_user() [116163.094850] null_blk: copy_from_nullb 1163 zero_user() [116163.094851] null_blk: copy_from_nullb 1163 zero_user() [116163.094852] null_blk: copy_from_nullb 1163 zero_user() [116163.094853] null_blk: copy_from_nullb 1163 zero_user() [116163.094854] null_blk: copy_from_nullb 1163 zero_user() [116163.094856] null_blk: copy_from_nullb 1163 zero_user() [116163.094857] null_blk: copy_from_nullb 1163 zero_user() [116163.094858] null_blk: copy_from_nullb 1163 zero_user() [116163.094859] null_blk: copy_from_nullb 1163 zero_user() [116163.094860] null_blk: copy_from_nullb 1163 zero_user() [116163.094861] null_blk: copy_from_nullb 1163 zero_user() [116163.094862] null_blk: copy_from_nullb 1163 zero_user() [116163.094863] null_blk: copy_from_nullb 1163 zero_user() [116163.094864] null_blk: copy_from_nullb 1163 zero_user() [116163.094865] null_blk: copy_from_nullb 1163 zero_user() [116163.094866] null_blk: copy_from_nullb 1163 zero_user() [116163.094867] null_blk: copy_from_nullb 1163 zero_user() [116163.094868] null_blk: copy_from_nullb 1163 zero_user() [116163.094869] null_blk: copy_from_nullb 1163 zero_user() [116163.094870] null_blk: copy_from_nullb 1163 zero_user() [116163.094871] null_blk: copy_from_nullb 1163 zero_user() [116163.094873] null_blk: copy_from_nullb 1163 zero_user() [116163.094874] null_blk: copy_from_nullb 1163 zero_user() [116163.094875] null_blk: copy_from_nullb 1163 zero_user() [116163.094876] null_blk: copy_from_nullb 1163 zero_user() [116163.094877] null_blk: copy_from_nullb 1163 zero_user() [116163.094878] null_blk: copy_from_nullb 1163 zero_user() [116163.094879] null_blk: copy_from_nullb 1163 zero_user() [116163.094880] null_blk: copy_from_nullb 1163 zero_user() linux-block (nullb-memcpy) # modprobe -r null_blk Debug Patch used for testing :-