Linux 5.10-rc1
Subsystems (2)
Two weeks have passed, and the merge window is over. I've tagged and pushed out 5.10-rc1, and everything looks fairly normal.
This looks to be a bigger release than I expected, and while the merge window is smaller than the one for 5.8 was, it's not a *lot* smaller. And 5.8 was our biggest release ever.
I'm not entirely sure whether this is just a general upward trend (we did seem to plateau for a while there), or just a fluke, or perhaps due to 5.9 dragging out an extra week. We will see, I guess.
That said, things seem to have gone fairly smoothly. I don't see any huge red flags, and the merge window didn't cause any unusual issues for me. Famous last words..
The most interesting - to me - change here is Christoph's setf_fs() removal (it got merged through Al Viro, as you can see in my mergelog below). It's not a _huge_ change, but it's interesting because the whole model of set_fs() to specify whether a userspace copy actually goes to user space or kernel space goes back to pretty much the original release of Linux, and while the name is entirely historic (it hasn't used the %fs segment register in a long time), the concept has remained. Until now.
We still do have "set_fs()" around, and not every architecture has been converted to the new world order, but x86, powerpc, s390 and RISC-V have had the address space overrides removed, and all the core work is done. Other architectures will hopefully get converted away from that very historic model too, but it might take a while to get rid of it all.
Anyway, to most people that all shouldn't matter at all, and it's mainly a small historical footnote that 5.10 no longer relies on the whole set_fs() model. Most of the actual changes are - as usual - driver updates, but there are changes all over. I think the merge log below gives some kind of flavor of what's been going on on a high level, but if you're interested in the details go look at the git tree. As mentioned, it's a big merge window, with almost 14k commits (*) by closer to 1700 people.
Please go test,
Linus
(*) closer to 15k commits if you count merges.
---
Al Viro (6):
copy_and_csum cleanups
compat iovec cleanups
compat quotactl cleanups
compat mount cleanups
initial set_fs() removal
misc vfs updates
Alex Williamson (1):
VFIO updates
Alexandre Belloni (1):
RTC updates
Andreas Gruenbacher (1):
gfs2 updates
Andrew Morton (3):
misc updates
more updates
yet more updates
Anna Schumaker (1):
NFS client updates
Arnaldo Carvalho de Melo (1):
perf tools updates
Benson Leung (1):
chrome platform updates
Bjorn Andersson (2):
remoteproc updates
rpmsg updates
Bjorn Helgaas (1):
PCI updates
Boris Brezillon (1):
i3c updates
Borislav Petkov (14):
EDAC updates
RAS updates
x86 cpu updates
x86 platform updates
x86 PASID updates
x86 fsgsbase updates
x86 fpu updates
x86 cleanups
x86 cache resource control updates
x86 fix
x86 fixes
x86 asm updates
x86 SEV-ES support
x86 SEV-ES fixes
Bruce Fields (1):
nfsd updates
Casey Schaufler (1):
smack updates
Christian Brauner (2):
kernel_clone() updates
pidfd updates
Christoph Hellwig (3):
dma-mapping updates
configfs updates
dma-mapping fixes
Corey Minyard (1):
IPMI updates
Damien Le Moal (1):
zonefs updates
Daniel Lezcano (1):
thermal updates
Daniel Thompson (1):
kgdb updates
Darrick Wong (5):
iomap updates
xfs updates
more xfs updates
clone/dedupe/remap code refactoring
xfs fixes
Dave Airlie (3):
drm updates
drm fixes
more drm fixes
David Howells (1):
afs updates
David Sterba (1):
btrfs updates
David Teigland (1):
dlm updates
Dmitry Torokhov (1):
input updates
Dominique Martinet (1):
9p updates
Eric Biggers (1):
fscrypt updates
Gao Xiang (1):
erofs updates
Geert Uytterhoeven (1):
m68k updates
Greg KH (6):
tty/serial updates
driver core updates
SPDX updates
staging / IIO driver updates
USB/PHY/Thunderbolt driver updates
char/misc driver updates
Greg Ungerer (1):
m68knommu updates
Guenter Roeck (1):
hwmon updates
Hans de Goede (1):
x86 platform driver updates
Helge Deller (2):
parisc updates
more parisc updates
Herbert Xu (1):
crypto updates
Ilya Dryomov (1):
ceph updates
Ingo Molnar (14):
scheduler updates
locking updates
EFI changes
orphan section checking
static call support
performance events updates
perf/kprobes updates
x86 kaslr updates
x86 mm updates
x86 build update
x86 paravirt cleanup
x86 Hyper-V update
objtool updates
RCU changes
Jaegeuk Kim (1):
f2fs updates
Jakub Kicinski (2):
networking updates
networking fixes
James Bottomley (2):
SCSI updates
more SCSI updates
Jan Kara (2):
UDF, reiserfs, ext2, quota fixes
direct-io fix
Jarkko Sakkinen (1):
tpm updates
Jason Gunthorpe (1):
rdma updates
Jassi Brar (1):
mailbox updates
Jean Delvare (1):
dmi update
Jeff Layton (1):
file locking fix
Jens Axboe (9):
block updates
io_uring updates
libata updates
block driver updates
io_uring updates
arch task_work cleanups
libata fixes
io_uring fixes
block fixes
Jessica Yu (1):
modules updates
Jiri Kosina (3):
livepatching update
HID updates
trivial updates
Joerg Roedel (2):
iommu updates
iommu fix
Jon Mason (1):
NTB fixes
Jonathan Corbet (2):
documentation updates
documentation fixes
Juergen Gross (3):
xen updates
more xen updates
more xen updates
Julia Lawall (1):
coccinelle updates
Kees Cook (2):
seccomp updates
overflow update
Konrad Rzeszutek Wilk (1):
swiotlb updates
Lee Jones (2):
MFD updates
backlight updates
Linus Walleij (2):
GPIO updates
pin control updates
Mark Brown (3):
regmap updates
regulator updates
spi updates
Masahiro Yamada (2):
Kbuild updates
Kconfig updates
Matthew Wilcox (1):
XArray updates
Mauro Carvalho Chehab (2):
media updates
documentation updates
Micah Morton (1):
SafeSetID updates
Michael Ellerman (2):
powerpc updates
powerpc fixes
Michael Tsirkin (1):
virtio updates
Michal Simek (1):
Microblaze build warning fix
Mike Snitzer (1):
device mapper updates
Miklos Szeredi (2):
overlayfs updates
fuse updates
Mimi Zohar (1):
integrity updates
Namjae Jeon (1):
exfat updates
Olof Johansson (5):
ARM SoC fixes
ARM SoC platform updates
ARM SoC-related driver updates
ARM Devicetree updates
ARM SoC defconfig updates
Palmer Dabbelt (2):
RISC-V updates
more RISC-V updates
Paolo Bonzini (2):
KVM updates
KVM fixes
Paul Moore (2):
audit updates
selinux updates
Pavel Machek (1):
LED updates
Petr Mladek (2):
printk updates
printk fix
Rafael Wysocki (5):
power management updates
ACPI updates
PNP updates
more power management updates
more ACPI updates
Richard Weinberger (4):
MTD updates
ubifs updates
more ubi and ubifs updates
UML updates
Rob Herring (1):
devicetree updates
Russell King (1):
ARM updates
Sebastian Reichel (1):
power supply and reset updates
Shuah Khan (4):
kselftest updates
kselftest updates
Kunit updates
more Kunit updates
Stephen Boyd (1):
clk updates
Steve French (2):
cifs updates
more cifs updates
Steven Rostedt (3):
tracing updates
tracing fix
tracing ring-buffer fix
Takashi Iwai (2):
sound updates
sound fixes
Ted Ts'o (1):
ext4 updates
Tejun Heo (1):
cgroup updates
Tetsuo HandaL
(1):
tomoyo fix
Thierry Reding (1):
pwm updates
Thomas Bogendoerfer (1):
MIPS updates
Thomas Gleixner (9):
debugobjects updates
timekeeping updates
irq updates
x86 irq updates
x86 entry code updates
locking fix
perf fix
scheduler fixes
timer fixes
Tony Luck (1):
ia64 updates
Ulf Hansson (1):
MMC updates
Vasily Gorbik (1):
s390 updates
Vineet Gupta (2):
ARC updates
ARC fix
Vinod Koul (1):
dmaengine updates
Wei Liu (2):
Hyper-V updates
another Hyper-V update
Will Deacon (2):
arm64 updates
more arm64 updates
Willy Tarreau (1):
random32 updates
Wim Van Sebroeck (1):
watchdog updates
Wolfram Sang (2):
i2c updates
i2c fix
linux-next: stats
Hi all,
As usual, the executive friendly graph is at
http://neuling.org/linux-next-size.html :-)
(No merge commits counted, next-20201009 was the first linux-next after the merge window opened.)
Commits in v5.10-rc1 (relative to v5.9): 13903 Commits in next-20201009: 13046 Commits with the same SHA1: 11911
Commits with the same patch_id: 563 (1)
Commits with the same subject line: 46 (1)
(1) not counting those in the lines above.
So commits in -rc1 that were in next-20201009: 12520 90%
Some breakdown of the list of extra commits (relative to next-20201009) in -rc1:
Top ten first word of commit summary:
158 kvm
112 perf
77 net
53 docs
43 io_uring
43 drm
32 bpf
31 dt-bindings
30 clk
28 nfsdTop ten authors:
77 [email protected]
64 [email protected]
28 [email protected]
20 [email protected]
19 [email protected]
18 [email protected]
17 [email protected]
17 [email protected]
16 [email protected]
16 [email protected]
Top ten commiters:
165 [email protected] 163 [email protected] 113 [email protected] 81 [email protected] 75 [email protected] 43 [email protected] 35 [email protected] 29 [email protected] 27 [email protected] 27 [email protected]
There are also 526 commits in next-20201009 that didn't make it into v5.10-rc1.
Top ten first word of commit summary:
81 drm
45 arm
32 tools
21 soc
19 bus
16 mm
13 dt-bindings
13 arm64
12 rcu
9 x86Top ten authors:
64 [email protected] 32 [email protected] 16 [email protected] 14 [email protected] 13 [email protected] 13 [email protected] 12 [email protected] 11 [email protected] 10 [email protected] 9 [email protected]
Some of Andrew's patches are fixes for other patches in his tree (and have been merged into those).
Top ten commiters:
103 [email protected] 83 [email protected] 71 [email protected] 26 [email protected] 19 [email protected] 17 [email protected] 15 [email protected] 14 [email protected] 14 [email protected] 13 [email protected]
Those commits by me are from Andrew's mmotm tree.
Signature · 2 lines
Cheers, Stephen Rothwell
problems with splice from /proc (was Linux 5.10-rc1)
In reply to Linus TorvaldsQuoted 8 lines · Linus Torvalds
The most interesting - to me - change here is Christoph's
setf_fs()removal (it got merged through Al Viro, as you can see in my mergelog below). It's not a _huge_ change, but it's interesting because the whole model ofset_fs()to specify whether a userspace copy actually goes to user space or kernel space goes back to pretty much the original release of Linux, and while the name is entirely historic (it hasn't used the %fs segment register in a long time), the concept has remained. Until now.
I told Al this yesterday, but figured I would mention it here for others to see.
Commit 36e2c7421f02 ("fs: don't allow splice read/write without explicit ops") from this patch series, is breaking the bionic test suite that does the following to verify that splice is working:
int in = open("/proc/cpuinfo", O_RDONLY);
ASSERT_NE(in, -1);
TemporaryFile tf;
ssize_t bytes_read = splice(in, nullptr, pipe_fds[1], nullptr, 8*1024, SPLICE_F_MORE | SPLICE_F_MOVE);
ASSERT_NE(bytes_read, -1);Before this change, all works well but now splice fails on /proc files (and I'm guessing other virtual filesystems).
I'll ask the bionic developers if they can change their test to some other file, but this is a regression and might show up in other "test platforms" as well. Using /proc for this is just so simple because these files are "always there" and don't require any housekeeping for test suites to worry about .
thanks,
greg k-h
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 24 lines · Greg KH
Quoted 8 lines · Linus Torvalds
The most interesting - to me - change here is Christoph's
setf_fs()removal (it got merged through Al Viro, as you can see in my mergelog below). It's not a _huge_ change, but it's interesting because the whole model ofset_fs()to specify whether a userspace copy actually goes to user space or kernel space goes back to pretty much the original release of Linux, and while the name is entirely historic (it hasn't used the %fs segment register in a long time), the concept has remained. Until now.I told Al this yesterday, but figured I would mention it here for others to see.
Commit
36e2c7421f02("fs: don't allow splice read/write without explicit ops") from this patch series, is breaking the bionic test suite that does the following to verify that splice is working:int in = open("/proc/cpuinfo", O_RDONLY); ASSERT_NE(in, -1); TemporaryFile tf; ssize_t bytes_read = splice(in, nullptr, pipe_fds[1], nullptr, 8*1024, SPLICE_F_MORE | SPLICE_F_MOVE); ASSERT_NE(bytes_read, -1);
Quoted 8 lines
Before this change, all works well but now splice fails on /proc files (and I'm guessing other virtual filesystems).
I'll ask the bionic developers if they can change their test to some other file, but this is a regression and might show up in other "test platforms" as well. Using /proc for this is just so simple because these files are "always there" and don't require any housekeeping for test suites to worry about .
Is this just a test or a real application? I already have the
infrastructure to support read_iter/write_iter on procfs and seq_files, but due to the intrusiveness we decided to only fix instances on an as needed basis. So we'll have everything ready once we pull the trigger.
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 34 lines · Christoph Hellwig
Quoted 24 lines · Greg KH
Quoted 8 lines · Linus Torvalds
The most interesting - to me - change here is Christoph's
setf_fs()removal (it got merged through Al Viro, as you can see in my mergelog below). It's not a _huge_ change, but it's interesting because the whole model ofset_fs()to specify whether a userspace copy actually goes to user space or kernel space goes back to pretty much the original release of Linux, and while the name is entirely historic (it hasn't used the %fs segment register in a long time), the concept has remained. Until now.I told Al this yesterday, but figured I would mention it here for others to see.
Commit
36e2c7421f02("fs: don't allow splice read/write without explicit ops") from this patch series, is breaking the bionic test suite that does the following to verify that splice is working:int in = open("/proc/cpuinfo", O_RDONLY); ASSERT_NE(in, -1); TemporaryFile tf; ssize_t bytes_read = splice(in, nullptr, pipe_fds[1], nullptr, 8*1024, SPLICE_F_MORE | SPLICE_F_MOVE); ASSERT_NE(bytes_read, -1);Quoted 8 lines
Before this change, all works well but now splice fails on /proc files (and I'm guessing other virtual filesystems).
I'll ask the bionic developers if they can change their test to some other file, but this is a regression and might show up in other "test platforms" as well. Using /proc for this is just so simple because these files are "always there" and don't require any housekeeping for test suites to worry about .
Quoted 4 lines
Is this just a test or a real application? I already have the
infrastructure to support read_iter/write_iter on procfs andseq_files, but due to the intrusiveness we decided to only fix instances on an as needed basis. So we'll have everything ready once we pull the trigger.
This is just a test, part of the bionic test suite to verify that bionic is working properly, and is run on new kernels as a verification that nothing functional broke in the kernel update.
I don't know about "real applications" yet.
Do you have to implement this on a per-proc-file-basis, or will it work for the whole filesystem?
And are the patches public anywhere that I could test them out?
thanks,
greg k-h
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 10 lines · Greg KH
This is just a test, part of the bionic test suite to verify that bionic is working properly, and is run on new kernels as a verification that nothing functional broke in the kernel update.
I don't know about "real applications" yet.
Do you have to implement this on a per-proc-file-basis, or will it work for the whole filesystem?
And are the patches public anywhere that I could test them out?
This all branch has the last posted version:
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-rw.2
with tthe proc:, sysctl: and seq_file: patches related to it. It did switch over all seq_file instances, but non-seq_file instances and write operations will need manual per-instance work.
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 12 lines · Christoph Hellwig
Quoted 10 lines · Greg KH
This is just a test, part of the bionic test suite to verify that bionic is working properly, and is run on new kernels as a verification that nothing functional broke in the kernel update.
I don't know about "real applications" yet.
Do you have to implement this on a per-proc-file-basis, or will it work for the whole filesystem?
And are the patches public anywhere that I could test them out?
Quoted 7 lines
This all branch has the last posted version:
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-rw.2
with tthe proc:, sysctl: and
seq_file: patches related to it. It did switch over allseq_fileinstances, but non-seq_file instances and write operations will need manual per-instance work.
Luckily /proc/cpuinfo seems to use the seq_file interface, so this series would work for that.
What's the odds of this series getting into 5.10-final? I'll go run it through the Android build system right now to see if it fixes the issue or not...
thanks,
greg k-h
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 6 lines · Greg KH
Luckily
/proc/cpuinfoseems to use theseq_fileinterface, so this series would work for that.What's the odds of this series getting into 5.10-final? I'll go run it through the Android build system right now to see if it fixes the issue or not...
It contained a few bits people didn't like too much, and we had no practical need for it. I'll try to respin it to be a little more palatable, and will update you once I have results.
Re: problems with splice from /proc (was Linux 5.10-rc1)
In reply to Greg KHQuoted 21 lines · Greg KH
Quoted 12 lines · Christoph Hellwig
Quoted 10 lines · Greg KH
This is just a test, part of the bionic test suite to verify that bionic is working properly, and is run on new kernels as a verification that nothing functional broke in the kernel update.
I don't know about "real applications" yet.
Do you have to implement this on a per-proc-file-basis, or will it work for the whole filesystem?
And are the patches public anywhere that I could test them out?
Quoted 7 lines
This all branch has the last posted version:
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-rw.2
with tthe proc:, sysctl: and
seq_file: patches related to it. It did switch over allseq_fileinstances, but non-seq_file instances and write operations will need manual per-instance work.
Quoted 6 lines
Luckily
/proc/cpuinfoseems to use theseq_fileinterface, so this series would work for that.What's the odds of this series getting into 5.10-final? I'll go run it through the Android build system right now to see if it fixes the issue or not...
Ok, I couldn't get a clean merge of that old branch on top of your 5.10-rc1 tree, so I can't give it a run-through. If you have an updated series you want me to test, I'll be glad to do so.
thanks,
greg k-h
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 3 lines · Greg KH
Ok, I couldn't get a clean merge of that old branch on top of your 5.10-rc1 tree, so I can't give it a run-through. If you have an updated series you want me to test, I'll be glad to do so.
Can you give this branch a spin?
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-proc
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 8 lines · Christoph Hellwig
Quoted 3 lines · Greg KH
Ok, I couldn't get a clean merge of that old branch on top of your 5.10-rc1 tree, so I can't give it a run-through. If you have an updated series you want me to test, I'll be glad to do so.
Can you give this branch a spin?
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-proc
Will do so in the morning, thanks!
greg k-h
Greg KH
Tested-byRe: problems with splice from /proc (was Linux 5.10-rc1)
In reply to Christoph HellwigQuoted 8 lines · Christoph Hellwig
Quoted 3 lines · Greg KH
Ok, I couldn't get a clean merge of that old branch on top of your 5.10-rc1 tree, so I can't give it a run-through. If you have an updated series you want me to test, I'll be glad to do so.
Can you give this branch a spin?
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-proc
Sorry for the delay, took a while to get results due to other testing errors...
Anyway, yes, this worked!
So feel free to add:
If you send those out.
Now I ran into a fnctl test that is failing when reading trying to run splice on /proc/version (crazy tests), let me go see if I can do the same thing you did for cpuinfo for all proc "single data" files...
thanks,
greg k-h
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 17 lines · Greg KH
Quoted 8 lines · Christoph Hellwig
Quoted 3 lines · Greg KH
Ok, I couldn't get a clean merge of that old branch on top of your 5.10-rc1 tree, so I can't give it a run-through. If you have an updated series you want me to test, I'll be glad to do so.
Can you give this branch a spin?
http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/set_fs-proc
Sorry for the delay, took a while to get results due to other testing errors...
Anyway, yes, this worked!
So feel free to add:
Quoted 7 lines
Tested-by Greg Kroah-Hartman <[email protected]>If you send those out.
Now I ran into a fnctl test that is failing when reading trying to run splice on
/proc/version(crazy tests), let me go see if I can do the same thing you did for cpuinfo for all proc "single data" files...
That worked too, I'll send a patch for this for the top of your branch to resolve this issue as a response to this email.
thanks,
greg k-h
Re: problems with splice from /proc (was Linux 5.10-rc1)
Quoted 2 lines · Greg KH
That worked too, I'll send a patch for this for the top of your branch to resolve this issue as a response to this email.
In the old branch I had a scripted conversion of all proc_read instances that point to seq_read to use seq_read_iter. I can probably do that as well here.
PATCH 1/1 proc "single files": switch to ->read_iter
In reply to Greg KHImplement ->read_iter for all proc "single files" so that more bionic tests cases can pass when they call splice() on other fun files like /proc/version
--- fs/proc/generic.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/proc/generic.c +1 -1
@@ -621,7 +621,7 @@ static int proc_single_open(struct inode *inode, struct file *file)static const struct proc_ops proc_single_ops = {/* not permanent -- can call into arbitrary ->single_show */ .proc_open = proc_single_open, - .proc_read = seq_read, + .proc_read_iter = seq_read_iter, .proc_lseek = seq_lseek, .proc_release = single_release, };
Signature · 1 line
2.29.1