warning for EOPNOTSUPP vfs_copy_file_range
Subsystems (3)
Hi,
We are experiencing the following warning from "WARN_ON_ONCE(ret == -EOPNOTSUPP);" in vfs_copy_file_range, from 64bf5ff58dff ("vfs: no fallback for ->copy_file_range")
# cat /sys/class/net/can0/phys_switch_id
WARNING: CPU: 7 PID: 673 at fs/read_write.c:1516 vfs_copy_file_range+0x380/0x440 Modules linked in: llce_can llce_logger llce_mailbox llce_core sch_fq_codel openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 CPU: 7 PID: 673 Comm: cat Not tainted 5.15.38-yocto-standard #1 Hardware name: Freescale S32G399A (DT) pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : vfs_copy_file_range+0x380/0x440 lr : vfs_copy_file_range+0x16c/0x440
sp : ffffffc00e0f3ce0
x29: ffffffc00e0f3ce0 x28: ffffff88157b5a40 x27: 0000000000000000
x26: ffffff8816ac3230 x25: ffffff881c060008 x24: 0000000000001000
x23: 0000000000000000 x22: 0000000000000000 x21: ffffff881cc99540
x20: ffffff881cc9a340 x19: ffffffffffffffa1 x18: ffffffffffffffff
x17: 0000000000000001 x16: 0000adfbb5178cde x15: ffffffc08e0f3647
x14: 0000000000000000 x13: 34613178302f3061 x12: 3178302b636e7973
x11: 0000000000058395 x10: 00000000fd1c5755 x9 : ffffffc008361950 x8 : ffffffc00a7d4d58 x7 : 0000000000000000 x6 : 0000000000000001 x5 : ffffffc009e81000 x4 : ffffffc009e817f8 x3 : 0000000000000000 x2 : 0000000000000000 x1 : ffffff88157b5a40 x0 : ffffffffffffffa1
Call trace: vfs_copy_file_range+0x380/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4
cat: /sys/class/net/can0/phys_switch_id: Operation not supported
And we found this is triggered by the following stack. Specifically, all netdev_ops in CAN drivers we can find now do not have ndo_get_port_parent_id and ndo_get_devlink_port, which makes phys_switch_id_show return -EOPNOTSUPP all the way back to vfs_copy_file_range.
phys_switch_id_show+0xf4/0x11c dev_attr_show+0x2c/0x6c sysfs_kf_seq_show+0xb8/0x150 kernfs_seq_show+0x38/0x44 seq_read_iter+0x1c4/0x4c0 kernfs_fop_read_iter+0x44/0x50 generic_file_splice_read+0xdc/0x190 do_splice_to+0xa0/0xfc splice_direct_to_actor+0xc4/0x250 do_splice_direct+0x94/0xe0 vfs_copy_file_range+0x16c/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4
According to the original commit log, this warning is for operational validity checks to generic_copy_file_range(). The reading will eventually return as not supported as printed above. But is this warning still necessary? If so we might want to remove it to have a cleaner dmesg.
Thanks,
Zhe
Quoted 65 lines · He Zhe
Hi,
We are experiencing the following warning from "
WARN_ON_ONCE(ret == -EOPNOTSUPP);" invfs_copy_file_range, from64bf5ff58dff("vfs: no fallback for ->copy_file_range")# cat /sys/class/net/can0/phys_switch_idWARNING: CPU: 7 PID: 673 at fs/read_write.c:1516 vfs_copy_file_range+0x380/0x440 Modules linked in: llce_can llce_logger llce_mailbox llce_core sch_fq_codel openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 CPU: 7 PID: 673 Comm: cat Not tainted 5.15.38-yocto-standard #1 Hardware name: Freescale S32G399A (DT) pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : vfs_copy_file_range+0x380/0x440 lr : vfs_copy_file_range+0x16c/0x440sp :
ffffffc00e0f3ce0
x29:ffffffc00e0f3ce0x28:ffffff88157b5a40x27: 0000000000000000
x26:ffffff8816ac3230x25:ffffff881c060008x24: 0000000000001000
x23: 0000000000000000 x22: 0000000000000000 x21:ffffff881cc99540
x20:ffffff881cc9a340x19:ffffffffffffffa1x18: ffffffffffffffff
x17: 0000000000000001 x16:0000adfbb5178cdex15:ffffffc08e0f3647
x14: 0000000000000000 x13:34613178302f3061x12:3178302b636e7973
x11: 0000000000058395 x10:00000000fd1c5755x9 :ffffffc008361950x8 :ffffffc00a7d4d58x7 : 0000000000000000 x6 : 0000000000000001 x5 :ffffffc009e81000x4 :ffffffc009e817f8x3 : 0000000000000000 x2 : 0000000000000000 x1 :ffffff88157b5a40x0 :ffffffffffffffa1Call trace: vfs_copy_file_range+0x380/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4cat:
/sys/class/net/can0/phys_switch_id: Operation not supportedAnd we found this is triggered by the following stack. Specifically, all
netdev_opsin CAN drivers we can find now do not havendo_get_port_parent_idandndo_get_devlink_port, which makesphys_switch_id_showreturn-EOPNOTSUPPall the way back tovfs_copy_file_range.phys_switch_id_show+0xf4/0x11c dev_attr_show+0x2c/0x6c sysfs_kf_seq_show+0xb8/0x150 kernfs_seq_show+0x38/0x44 seq_read_iter+0x1c4/0x4c0 kernfs_fop_read_iter+0x44/0x50 generic_file_splice_read+0xdc/0x190 do_splice_to+0xa0/0xfc splice_direct_to_actor+0xc4/0x250 do_splice_direct+0x94/0xe0 vfs_copy_file_range+0x16c/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4
Quoted 5 lines
According to the original commit log, this warning is for operational validity checks to
generic_copy_file_range(). The reading will eventually return as not supported as printed above. But is this warning still necessary? If so we might want to remove it to have a cleaner dmesg.
Sigh! Those filesystems have no business doing copy_file_range()
Here is a patch that Luis has been trying to push last year to fix a problem with copy_file_range() from tracefs:
https://lore.kernel.org/linux-fsdevel/[email protected]/
Luis gave up on it, because no maintainer stepped up to take the patch, but I think that is the right way to go.
Maybe this bug report can raise awareness to that old patch.
Al, could you have a look?
Thanks,
Amir.
Quoted 72 lines · Amir Goldstein
Quoted 65 lines · He Zhe
Hi,
We are experiencing the following warning from "
WARN_ON_ONCE(ret == -EOPNOTSUPP);" invfs_copy_file_range, from64bf5ff58dff("vfs: no fallback for ->copy_file_range")# cat /sys/class/net/can0/phys_switch_idWARNING: CPU: 7 PID: 673 at fs/read_write.c:1516 vfs_copy_file_range+0x380/0x440 Modules linked in: llce_can llce_logger llce_mailbox llce_core sch_fq_codel openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 CPU: 7 PID: 673 Comm: cat Not tainted 5.15.38-yocto-standard #1 Hardware name: Freescale S32G399A (DT) pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : vfs_copy_file_range+0x380/0x440 lr : vfs_copy_file_range+0x16c/0x440sp :
ffffffc00e0f3ce0
x29:ffffffc00e0f3ce0x28:ffffff88157b5a40x27: 0000000000000000
x26:ffffff8816ac3230x25:ffffff881c060008x24: 0000000000001000
x23: 0000000000000000 x22: 0000000000000000 x21:ffffff881cc99540
x20:ffffff881cc9a340x19:ffffffffffffffa1x18: ffffffffffffffff
x17: 0000000000000001 x16:0000adfbb5178cdex15:ffffffc08e0f3647
x14: 0000000000000000 x13:34613178302f3061x12:3178302b636e7973
x11: 0000000000058395 x10:00000000fd1c5755x9 :ffffffc008361950x8 :ffffffc00a7d4d58x7 : 0000000000000000 x6 : 0000000000000001 x5 :ffffffc009e81000x4 :ffffffc009e817f8x3 : 0000000000000000 x2 : 0000000000000000 x1 :ffffff88157b5a40x0 :ffffffffffffffa1Call trace: vfs_copy_file_range+0x380/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4cat:
/sys/class/net/can0/phys_switch_id: Operation not supportedAnd we found this is triggered by the following stack. Specifically, all
netdev_opsin CAN drivers we can find now do not havendo_get_port_parent_idandndo_get_devlink_port, which makesphys_switch_id_showreturn-EOPNOTSUPPall the way back tovfs_copy_file_range.phys_switch_id_show+0xf4/0x11c dev_attr_show+0x2c/0x6c sysfs_kf_seq_show+0xb8/0x150 kernfs_seq_show+0x38/0x44 seq_read_iter+0x1c4/0x4c0 kernfs_fop_read_iter+0x44/0x50 generic_file_splice_read+0xdc/0x190 do_splice_to+0xa0/0xfc splice_direct_to_actor+0xc4/0x250 do_splice_direct+0x94/0xe0 vfs_copy_file_range+0x16c/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4Quoted 5 lines
According to the original commit log, this warning is for operational validity checks to
generic_copy_file_range(). The reading will eventually return as not supported as printed above. But is this warning still necessary? If so we might want to remove it to have a cleaner dmesg.
Quoted 6 lines
Sigh! Those filesystems have no business doing
copy_file_range()Here is a patch that Luis has been trying to push last year to fix a problem with
copy_file_range()from tracefs:
Yikes! It's been a while and I completely forgot about it. I can definitely try to respin this patch if someone's interested in picking it. I'll have to go re-read everything again and see what's missing and what has changed in between.
Cheers,
Signature · 1 line
Luís
Quoted 10 lines
Luis gave up on it, because no maintainer stepped up to take the patch, but I think that is the right way to go.
Maybe this bug report can raise awareness to that old patch.
Al, could you have a look?
Thanks,
Amir.
Quoted 74 lines · Luís Henriques
Quoted 72 lines · Amir Goldstein
Quoted 64 lines · He Zhe
Hi,
We are experiencing the following warning from "
WARN_ON_ONCE(ret == -EOPNOTSUPP);" invfs_copy_file_range, from64bf5ff58dff("vfs: no fallback for ->copy_file_range")# cat /sys/class/net/can0/phys_switch_idWARNING: CPU: 7 PID: 673 at fs/read_write.c:1516 vfs_copy_file_range+0x380/0x440 Modules linked in: llce_can llce_logger llce_mailbox llce_core sch_fq_codel openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 CPU: 7 PID: 673 Comm: cat Not tainted 5.15.38-yocto-standard #1 Hardware name: Freescale S32G399A (DT) pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : vfs_copy_file_range+0x380/0x440 lr : vfs_copy_file_range+0x16c/0x440sp :
ffffffc00e0f3ce0
x29:ffffffc00e0f3ce0x28:ffffff88157b5a40x27: 0000000000000000
x26:ffffff8816ac3230x25:ffffff881c060008x24: 0000000000001000
x23: 0000000000000000 x22: 0000000000000000 x21:ffffff881cc99540
x20:ffffff881cc9a340x19:ffffffffffffffa1x18: ffffffffffffffff
x17: 0000000000000001 x16:0000adfbb5178cdex15:ffffffc08e0f3647
x14: 0000000000000000 x13:34613178302f3061x12:3178302b636e7973
x11: 0000000000058395 x10:00000000fd1c5755x9 :ffffffc008361950x8 :ffffffc00a7d4d58x7 : 0000000000000000 x6 : 0000000000000001 x5 :ffffffc009e81000x4 :ffffffc009e817f8x3 : 0000000000000000 x2 : 0000000000000000 x1 :ffffff88157b5a40x0 :ffffffffffffffa1Call trace: vfs_copy_file_range+0x380/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4cat:
/sys/class/net/can0/phys_switch_id: Operation not supportedAnd we found this is triggered by the following stack. Specifically, all
netdev_opsin CAN drivers we can find now do not havendo_get_port_parent_idandndo_get_devlink_port, which makesphys_switch_id_showreturn-EOPNOTSUPPall the way back tovfs_copy_file_range.phys_switch_id_show+0xf4/0x11c dev_attr_show+0x2c/0x6c sysfs_kf_seq_show+0xb8/0x150 kernfs_seq_show+0x38/0x44 seq_read_iter+0x1c4/0x4c0 kernfs_fop_read_iter+0x44/0x50 generic_file_splice_read+0xdc/0x190 do_splice_to+0xa0/0xfc splice_direct_to_actor+0xc4/0x250 do_splice_direct+0x94/0xe0 vfs_copy_file_range+0x16c/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4Quoted 5 lines
According to the original commit log, this warning is for operational validity checks to
generic_copy_file_range(). The reading will eventually return as not supported as printed above. But is this warning still necessary? If so we might want to remove it to have a cleaner dmesg.Sigh! Those filesystems have no business doing
copy_file_range()
Quoted 8 lines
Quoted 4 lines
Here is a patch that Luis has been trying to push last year to fix a problem with
copy_file_range()from tracefs:Yikes! It's been a while and I completely forgot about it. I can definitely try to respin this patch if someone's interested in picking it. I'll have to go re-read everything again and see what's missing and what has changed in between.
Thank you both for quick replies.
It would be good if this could be sorted out, as folks who are not familiar with it might be confused by the call trace. But if this is supposed to cost a long time, maybe we can first solve the false positive warning for the drivers in this case, as it seems the "operational validity checks" was not for these drivers.
Regards,
Zhe
Quoted 2 lines
Cheers,
Quoted 87 lines · He Zhe
Quoted 74 lines · Luís Henriques
Quoted 72 lines · Amir Goldstein
Quoted 64 lines · He Zhe
Hi,
We are experiencing the following warning from "
WARN_ON_ONCE(ret == -EOPNOTSUPP);" invfs_copy_file_range, from64bf5ff58dff("vfs: no fallback for ->copy_file_range")# cat /sys/class/net/can0/phys_switch_idWARNING: CPU: 7 PID: 673 at fs/read_write.c:1516 vfs_copy_file_range+0x380/0x440 Modules linked in: llce_can llce_logger llce_mailbox llce_core sch_fq_codel openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 CPU: 7 PID: 673 Comm: cat Not tainted 5.15.38-yocto-standard #1 Hardware name: Freescale S32G399A (DT) pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) pc : vfs_copy_file_range+0x380/0x440 lr : vfs_copy_file_range+0x16c/0x440sp :
ffffffc00e0f3ce0
x29:ffffffc00e0f3ce0x28:ffffff88157b5a40x27: 0000000000000000
x26:ffffff8816ac3230x25:ffffff881c060008x24: 0000000000001000
x23: 0000000000000000 x22: 0000000000000000 x21:ffffff881cc99540
x20:ffffff881cc9a340x19:ffffffffffffffa1x18: ffffffffffffffff
x17: 0000000000000001 x16:0000adfbb5178cdex15:ffffffc08e0f3647
x14: 0000000000000000 x13:34613178302f3061x12:3178302b636e7973
x11: 0000000000058395 x10:00000000fd1c5755x9 :ffffffc008361950x8 :ffffffc00a7d4d58x7 : 0000000000000000 x6 : 0000000000000001 x5 :ffffffc009e81000x4 :ffffffc009e817f8x3 : 0000000000000000 x2 : 0000000000000000 x1 :ffffff88157b5a40x0 :ffffffffffffffa1Call trace: vfs_copy_file_range+0x380/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4cat:
/sys/class/net/can0/phys_switch_id: Operation not supportedAnd we found this is triggered by the following stack. Specifically, all
netdev_opsin CAN drivers we can find now do not havendo_get_port_parent_idandndo_get_devlink_port, which makesphys_switch_id_showreturn-EOPNOTSUPPall the way back tovfs_copy_file_range.phys_switch_id_show+0xf4/0x11c dev_attr_show+0x2c/0x6c sysfs_kf_seq_show+0xb8/0x150 kernfs_seq_show+0x38/0x44 seq_read_iter+0x1c4/0x4c0 kernfs_fop_read_iter+0x44/0x50 generic_file_splice_read+0xdc/0x190 do_splice_to+0xa0/0xfc splice_direct_to_actor+0xc4/0x250 do_splice_direct+0x94/0xe0 vfs_copy_file_range+0x16c/0x440 __do_sys_copy_file_range+0x178/0x3a4 __arm64_sys_copy_file_range+0x34/0x4c invoke_syscall+0x5c/0x130 el0_svc_common.constprop.0+0x68/0x124 do_el0_svc+0x50/0xbc el0_svc+0x54/0x130 el0t_64_sync_handler+0xa4/0x130 el0t_64_sync+0x1a0/0x1a4Quoted 5 lines
According to the original commit log, this warning is for operational validity checks to
generic_copy_file_range(). The reading will eventually return as not supported as printed above. But is this warning still necessary? If so we might want to remove it to have a cleaner dmesg.Sigh! Those filesystems have no business doing
copy_file_range()Quoted 8 lines
Quoted 4 lines
Here is a patch that Luis has been trying to push last year to fix a problem with
copy_file_range()from tracefs:Yikes! It's been a while and I completely forgot about it. I can definitely try to respin this patch if someone's interested in picking it. I'll have to go re-read everything again and see what's missing and what has changed in between.
Quoted 7 lines
Thank you both for quick replies.
It would be good if this could be sorted out, as folks who are not familiar with it might be confused by the call trace. But if this is supposed to cost a long time, maybe we can first solve the false positive warning for the drivers in this case, as it seems the "operational validity checks" was not for these drivers.
Yes, technically, you are right.
Userspace should not be able to trigger a code validity assertion. But the reason that assertion is there is to warn us developers if we had overlooked a logic case and IMO we did.
The entire concept of calling ->copy_file_range() on random filesystems has more than one problem and I would like for the kernel to stop doing that.
Thanks,
Amir.
Quoted 13 lines · [Re: warning for EOPNOTSUPP vfs_copy_file_range] On 19/05/2022 (Thu 15:31)…
Quoted 11 lines · Amir Goldstein
Quoted 10 lines · He Zhe
Hi,
We are experiencing the following warning from "
WARN_ON_ONCE(ret == -EOPNOTSUPP);" invfs_copy_file_range, from64bf5ff58dff("vfs: no fallback for ->copy_file_range")# cat /sys/class/net/can0/phys_switch_idWARNING: CPU: 7 PID: 673 at fs/read_write.c:1516 vfs_copy_file_range+0x380/0x440
[...]
Quoted 11 lines
Quoted 6 lines
Sigh! Those filesystems have no business doing
copy_file_range()Here is a patch that Luis has been trying to push last year to fix a problem with
copy_file_range()from tracefs:Yikes! It's been a while and I completely forgot about it. I can definitely try to respin this patch if someone's interested in picking it. I'll have to go re-read everything again and see what's missing and what has changed in between.
If it helps any, you don't need CAN bus or anything complicated. I was able to reproduce it with the loopback device on linux-next of today with a new userspace (from Yocto) that actively uses copy_file_range()
Note that a cp or redirect is needed to trigger the copy - with the new userspace requiremet. Or write your own reproducer that goes at the syscall directly on an old userspace (untested by me).
root@qemux86-64:/sys/class/net/lo# cat phys_switch_id cat: phys_switch_id: Operation not supported root@qemux86-64:/sys/class/net/lo# cat phys_switch_id > /tmp/foo [ 87.527506] ------------[ cut here ]------------ [ 87.527675] WARNING: CPU: 2 PID: 238 at /home/paul/git/linux-head/fs/read_write.c:1511 vfs_copy_file_range+0x47c/0x4e0
Paul.
--
Quoted 15 lines
Cheers,
Signature · 1 line
LuísQuoted 10 lines
Luis gave up on it, because no maintainer stepped up to take the patch, but I think that is the right way to go.
Maybe this bug report can raise awareness to that old patch.
Al, could you have a look?
Thanks,
Amir.
Al Viro
maintainerQuoted 6 lines · Amir Goldstein
Luis gave up on it, because no maintainer stepped up to take the patch, but I think that is the right way to go.
Maybe this bug report can raise awareness to that old patch.
Al, could you have a look?
IIRC, you had objections to that variant back then...
Quoted 11 lines · Al Viro
Quoted 6 lines · Amir Goldstein
Luis gave up on it, because no maintainer stepped up to take the patch, but I think that is the right way to go.
Maybe this bug report can raise awareness to that old patch.
Al, could you have a look?
IIRC, you had objections to that variant back then...
Right. But not about the "main" patch.
The patch had an "also" part:
The short-circuit code for the case where the copy length is zero has also been dropped from the VFS code. This is because a zero size copy between two files shall provide a clear indication on whether or not the filesystem supports non-zero copies.
- if (len == 0)
- return 0;
-
Which would have been a regression for nfs client, because
nfs protocol treats length 0 from ->copy_file_range() as "copy everything":
This api impedance should be fixed in the nfs client, but I'm not sure if that was already done.
I will test and re-post Luis' patch without removing the short-circuit unless Luis gets to it first.
BTW, IIRC, there were already LTP tests and man page fixes posted for the copt_file_range() behavior change.
Thanks,
Amir.
Quoted 29 lines · Amir Goldstein
Quoted 11 lines · Al Viro
Quoted 6 lines · Amir Goldstein
Luis gave up on it, because no maintainer stepped up to take the patch, but I think that is the right way to go.
Maybe this bug report can raise awareness to that old patch.
Al, could you have a look?
IIRC, you had objections to that variant back then...
Right. But not about the "main" patch.
The patch had an "also" part:The short-circuit code for the case where the copy length is zero has also been dropped from the VFS code. This is because a zero size copy between two files shall provide a clear indication on whether or not the filesystem supports non-zero copies.
(quoted diff) +0
-3- if (len == 0) - return 0; -Which would have been a regression for nfs client, because
nfs protocol treats length 0 from ->copy_file_range() as "copy everything":
Quoted 8 lines
This api impedance should be fixed in the nfs client, but I'm not sure if that was already done.
I will test and re-post Luis' patch without removing the short-circuit unless Luis gets to it first.
Urgh! That old patch passes the fstests -g copy_range group on nfs, but fails almost all of them on xfs/btrfs.
The reason is that when we allow to perform copy_range with remap_file_range() it fails for sizes smaller than block size and returning short read of 0 from copy_range is not an option.
So what I am going to do is to keep the basic restriction in this patch of:
"copy_range allowed for fs that implement either ->copy_file_range() or ->remap_file_range() (for same sb copy)"
But will change the logic of:
"try clone and then copy then fail"
to:
- try ->copy_file_range()
- try ->remap_file_range()
- fall back to kernel copy
Patch coming shortly.
Thanks,
Amir.