m68k uaccess fault handling fixes
v4 cover m68k uaccess fault handling fixes
(Actually now) version 4 of fixes for uaccess fault handling on 68030, tested on 68030 and 68040.
Patch 1 has one exception table entry fixed (residual calculated in __generic_copy_to_user did not take into account the number of longword transfers omitted due to a fault) si as well as a final NOP added in __clear_user.
Patch 2 is unchanged from v3 (actually since RFC v1), now has Tested-by tag by Finn Thain added.
These patches would still benefit from testing on 68060 and Coldfire.
Cheers,
Michael
PATCH v4 1/2 m68k: Handle __generic_copy_to_user faults more carefully
As mentioned by Finn Thain in his patch to improve put_user exception handling on 040, a similar problem exists on 030 processors.
A moves instruction that crosses a page boundary from a mapped page into an unmapped one will cause a mid-instruction bus error exception (frame format b), with the PC pointing (usually) two instructions past the faulting movesl instruction.
Our exception handling in __generic_copy_to_user only covers the instruction immediately following the faulting one. As a result, fixup_exception in send_fault_sig does not detect this case, and cause send_fault_sig to oops.
Extend the exception table to cover one additional instruction beyond the moves[lwb] instructions.
Tested on 68030 (Atari Falcon 030) with transfers beginning at one to six bytes offset from the end of a mapped page, followed by further bytes on an unmapped page (testcase derived from stress-ng sysbadaddr stressor by Finn Thain).
Tested on 68040 (Mac Quadra) and 68030 (Mac IIci) by Finn Thain.
A similar problem is present in __clear_user(); modify the exception table for that function in the same way (tested by Finn Thain).
---
Changes from v3:
Finn Thain:
- correct exception table entry for movew instruction in __generic_copy_to_user
- add final NOP in __clear_user
Changes from RFC v2:
Finn Thain:
- add missing extension table entries and final NOP in __generic_copy_to_user faults in 040 tests
Michael Schmitz:
- add yet another exception table entry in __clear_user
Changes from RFC v1:
Michael Schmitz:
- use extended exception table instead of additional NOPs
--- arch/m68k/lib/uaccess.c | 65 ++++++++++++++++++++++++----------------- 1 file changed, 38 insertions(+), 27 deletions(-)
arch/m68k/lib/uaccess.c +38 -27
@@ -62,35 +62,42 @@ unsigned long __generic_copy_to_user(void __user *to, const void *from, asmvolatile ("\n" " tst.l %0\n" -" jeq 4f\n" +" jeq 5f\n" "1: move.l (%1)+,%3\n" "2: " MOVES ".l %3,(%2)+\n" "3: subq.l #1,%0\n" -" jne 1b\n" -"4: btst #1,%5\n" -" jeq 6f\n" -" move.w (%1)+,%3\n" -"5: " MOVES ".w %3,(%2)+\n" -"6: btst #0,%5\n" +"4: jne 1b\n" +"5: btst #1,%5\n" " jeq 8f\n" -" move.b (%1)+,%3\n" -"7: " MOVES ".b %3,(%2)+\n" -"8:\n" +"6: move.w (%1)+,%3\n" +"7: " MOVES ".w %3,(%2)+\n" +"8: btst #0,%5\n" +"9: jeq 13f\n" +"10: move.b (%1)+,%3\n" +"11: " MOVES ".b %3,(%2)+\n" +"12: nop\n" +"13:\n" " .section .fixup,\"ax\"\n" " .even\n" "20: lsl.l #2,%0\n" "50: add.l %5,%0\n" -" jra 8b\n" +" jra 13b\n" " .previous\n" "\n" " .section __ex_table,\"a\"\n" " .align 4\n" +" .long 1b,20b\n" " .long 2b,20b\n" " .long 3b,20b\n" -" .long 5b,50b\n" -" .long 6b,50b\n" +" .long 4b,20b\n" +" .long 5b,20b\n" +" .long 6b,20b\n" " .long 7b,50b\n" " .long 8b,50b\n" +" .long 9b,50b\n" +" .long 10b,50b\n" +" .long 11b,50b\n" +" .long 12b,50b\n" " .previous" :"=d" (res),"+a" (from),"+a" (to),"=&d" (tmp) :"0" (n /4 ),"d" (n &3 ));
@@ -109,32 +116,36 @@ unsigned long __clear_user(void __user *to, unsigned long n) asmvolatile ("\n" " tst.l %0\n" -" jeq 3f\n" +" jeq 4f\n" "1: " MOVES ".l %2,(%1)+\n" "2: subq.l #1,%0\n" -" jne 1b\n" -"3: btst #1,%4\n" -" jeq 5f\n" -"4: " MOVES ".w %2,(%1)+\n" -"5: btst #0,%4\n" -" jeq 7f\n" -"6: " MOVES ".b %2,(%1)\n" -"7:\n" +"3: jne 1b\n" +"4: btst #1,%4\n" +" jeq 6f\n" +"5: " MOVES ".w %2,(%1)+\n" +"6: btst #0,%4\n" +"7: jeq 9f\n" +"8: " MOVES ".b %2,(%1)\n" +"9: nop\n" +"10:\n" " .section .fixup,\"ax\"\n" " .even\n" -"10: lsl.l #2,%0\n" +"20: lsl.l #2,%0\n" "40: add.l %4,%0\n" -" jra 7b\n" +" jra 10b\n" " .previous\n" "\n" " .section __ex_table,\"a\"\n" " .align 4\n" -" .long 1b,10b\n" -" .long 2b,10b\n" -" .long 4b,40b\n" +" .long 1b,20b\n" +" .long 2b,20b\n" +" .long 3b,20b\n" +" .long 4b,20b\n" " .long 5b,40b\n" " .long 6b,40b\n" " .long 7b,40b\n" +" .long 8b,40b\n" +" .long 9b,40b\n" " .previous" :"=d" (res),"+a" (to) :"d" (0 ),"0" (n /4 ),"d" (n &3 ));
Signature · 1 line
2.17.1
PATCH v4 2/2 m68k: improve __constant_copy_to_user_asm() fault handling
In reply to Michael SchmitzA problem similar to that reported for __put_user_asm and __generic_copy_to_user is also present in __constant_copy_to_user_asm.
Address the problem by extending the exception table to cover two instructions past each moves instruction, and adding a single NOP at the very end to catch faults on the final instruction (which is not guaranteed to be a movesb!).
Tested on 68030 (Atari Falcon 030) with a transfer beginning at a single byte at the end of a mapped page followed by seven more bytes on an unmapped page (testcase derived from stress-ng sysbadaddr stressor by Finn Thain and modified to use the llseek syscall).
--- arch/m68k/include/asm/uaccess.h | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
arch/m68k/include/asm/uaccess.h +6 -3
@@ -288,10 +288,11 @@ __constant_copy_from_user(void *to, const void __user *from, unsigned long n) "21: "MOVES"."#s2" %3,(%1)+\n" \ "22:\n" \ " .ifnc \""#s3"\",\"\"\n" \ - " move."#s3" (%2)+,%3\n" \ - "31: "MOVES"."#s3" %3,(%1)+\n" \ - "32:\n" \ + "31: move."#s3" (%2)+,%3\n" \ + "32: "MOVES"."#s3" %3,(%1)+\n" \ + "33:\n" \ " .endif\n" \ + "34: nop\n" \ "4:\n" \ "\n" \ " .section __ex_table,\"a\"\n" \
@@ -303,7 +304,9 @@ __constant_copy_from_user(void *to, const void __user *from, unsigned long n) " .ifnc \""#s3"\",\"\"\n" \" .long 31b,5f\n" \" .long 32b,5f\n" \ +" .long 33b,5f\n" \" .endif\n" \ +" .long 34b,5f\n" \" .previous\n" \"\n" \" .section .fixup,\"ax\"\n" \
Signature · 1 line
2.17.1
Hi Michael,
Quoted 8 lines · Michael Schmitz
(Actually now) version 4 of fixes for uaccess fault handling on 68030, tested on 68030 and 68040.
Patch 1 has one exception table entry fixed (residual calculated in
__generic_copy_to_userdid not take into account the number of longword transfers omitted due to a fault) si as well as a final NOP added in__clear_user.
Quoted 5 lines
Patch 2 is unchanged from v3 (actually since RFC v1), now has Tested-by tag by Finn Thain added.
These patches would still benefit from testing on 68060 and Coldfire.
I would very much like to test them on ColdFire. Unfortunately my only ColdFire with MMU board (a M547xEVB) failed a couple of months back. I am trying to find a replacement, but so far have not been able to get my hands on one.
Maybe someone else out there has one they can test with?
Regards
Greg
Geert Uytterhoeven
maintainerQuoted 15 lines · Greg Ungerer
Quoted 8 lines · Michael Schmitz
(Actually now) version 4 of fixes for uaccess fault handling on 68030, tested on 68030 and 68040.
Patch 1 has one exception table entry fixed (residual calculated in
__generic_copy_to_userdid not take into account the number of longword transfers omitted due to a fault) si as well as a final NOP added in__clear_user.Quoted 5 lines
Patch 2 is unchanged from v3 (actually since RFC v1), now has Tested-by tag by Finn Thain added.
These patches would still benefit from testing on 68060 and Coldfire.
Quoted 6 lines
I would very much like to test them on ColdFire. Unfortunately my only ColdFire with MMU board (a M547xEVB) failed a couple of months back. I am trying to find a replacement, but so far have not been able to get my hands on one.
Maybe someone else out there has one they can test with?
Angelo?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
In reply to Michael SchmitzQuoted 22 lines · Michael Schmitz
As mentioned by Finn Thain in his patch to improve
put_userexception handling on 040, a similar problem exists on 030 processors.A moves instruction that crosses a page boundary from a mapped page into an unmapped one will cause a mid-instruction bus error exception (frame format b), with the PC pointing (usually) two instructions past the faulting movesl instruction.
Our exception handling in
__generic_copy_to_useronly covers the instruction immediately following the faulting one. As a result,fixup_exceptioninsend_fault_sigdoes not detect this case, and causesend_fault_sigto oops.Extend the exception table to cover one additional instruction beyond the moves[lwb] instructions.
Tested on 68030 (Atari Falcon 030) with transfers beginning at one to six bytes offset from the end of a mapped page, followed by further bytes on an unmapped page (testcase derived from stress-ng sysbadaddr stressor by Finn Thain).
Quoted 6 lines
Tested on 68040 (Mac Quadra) and 68030 (Mac IIci) by Finn Thain.
A similar problem is present in
__clear_user(); modify the exception table for that function in the same way (tested by Finn Thain).
Well, that __clear_user() bug is no longer theoretical. I accidentally bumped into it when I sent a ^C to a shell script I wrote to test some mac_scsi driver patches...
[ 0.000000] Linux version 6.10.0-mac-00011-gd1d490afeb2a (fthain@nippy) (m68k-unknown-linux-musl-gcc (Gentoo 13.2.1_p20240210 p14) 13.2.1 20240210, GNU ld (Gentoo 2.40 p7) 2.40.0) #39 Sat Aug 3 19:57:43 AEST 2024
...
root@(none):/# bash scsi-test.sh /dev/sda5 bs=512k count=1 [ 130.090000] sd 0:0:0:0: [sda] tag#9 PDMA fixup: DRQ timeout [ 130.090000] sd 0:0:0:0: [sda] tag#9 switching to slow handshake [ 130.180000] sd 0:0:0:0: Power-on or device reset occurred 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 5.13732 s, 102 kB/s [ 136.030000] bash (57): drop_caches: 3 ^C root@(none):/# bash scsi-test.sh /dev/sdb5 bs=512k count=1 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.1273 s, 465 kB/s [ 160.440000] bash (61): drop_caches: 3 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.12763 s, 465 kB/s [ 171.920000] bash (61): drop_caches: 3
bs=64k count=8
8+0 records in
8+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.03469 s, 507 kB/s
[ 184.480000] bash (61): drop_caches: 3 8+0 records in 8+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.03221 s, 508 kB/s [ 196.340000] bash (61): drop_caches: 3
bs=4k count=128
128+0 records in
128+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.13089 s, 464 kB/s
[ 208.860000] bash (61): drop_caches: 3 128+0 records in 128+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.13089 s, 464 kB/s [ 220.850000] bash (61): drop_caches: 3
bs=512 count=1k
1024+0 records in
1024+0 records out
524288 bytes (524 kB, 512 KiB) copied, 3.69678 s, 142 kB/s
[ 236.020000] bash (61): drop_caches: 3 1024+0 records in 1024+0 records out 524288 bytes (524 kB, 512 KiB) copied, 3.68874 s, 142 kB/s [ 250.570000] bash (61): drop_caches: 3
bs=512k count=1
1+0 records in
1+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.12338 s, 467 kB/s
[ 263.120000] bash (61): drop_caches: 3 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.12665 s, 465 kB/s [ 275.110000] bash (61): drop_caches: 3
bs=64k count=8
8+0 records in
8+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.08861 s, 482 kB/s
[ 287.610000] bash (61): drop_caches: 3 8+0 records in 8+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.03319 s, 507 kB/s [ 299.510000] bash (61): drop_caches: 3 ^C[ 303.190000] Unable to handle kernel access at virtual address d7d17fd7 [ 303.200000] Oops: 00000000 [ 303.210000] Modules linked in: [ 303.220000] PC: [<004301fe>] __clear_user+0x22/0x40 [ 303.240000] SR: 2000 SP: 4841d159 a2: 00b78590 [ 303.250000] d0: 000003ab d1: 00000000 d2: 00000000 d3: 8000c150 [ 303.260000] d4: 00002000 d5: 00000000 a0: 8000c154 a1: 009bdb3c [ 303.280000] Process cmp (pid: 92, task=7d95deea) [ 303.290000] Frame format=B ssw=0709 isc=0801 isb=0001 daddr=8000c150 dobuf=00000000 [ 303.310000] baddr=00430202 dibuf=8000c150 ver=f [ 303.320000] Stack from 0149de8c: [ 303.320000] 8001c2c4 0149deb8 00139904 8000c150 00000eb0 00983c00 00000000 00000003 [ 303.320000] 00000003 8000bf00 00983a00 0149df40 0013a1fa 00912480 8000bf00 00983c60 [ 303.320000] 00000003 00000012 00000000 fffffff8 00000006 d012f728 00000000 00983a5a [ 303.320000] 00000004 00983a00 000579d4 0052b7e8 009125a0 00912480 00983a5a 80000034 [ 303.320000] 000d1704 014994c0 80008864 80000000 00000000 80008864 80008864 00000000 [ 303.320000] 01497800 00000000 00000034 fffffff8 fffffff8 0149df7c 000e9296 00983a00 [ 303.380000] Call Trace: [<00139904>] elf_load+0x192/0x1da [ 303.390000] [<0013a1fa>] load_elf_binary+0x8ae/0xe46 [ 303.400000] [<000579d4>] try_module_get+0x0/0x48 [ 303.410000] [<000d1704>] kfree+0x0/0x160 [ 303.420000] [<000e9296>] bprm_execve+0x134/0x416 [ 303.430000] [<000e9daa>] copy_string_kernel+0x0/0x184 [ 303.440000] [<000e9fa4>] copy_strings+0x0/0x310 [ 303.460000] [<000ea3f6>] do_execveat_common+0x142/0x1d2 [ 303.480000] [<000eb128>] sys_execve+0x2a/0x36 [ 303.490000] [<0000275e>] syscall+0x8/0xc [ 303.510000] [<0008c010>] do_read_cache_folio+0xde/0x2c8 [ 303.520000] [ 303.520000] Code: 206e 0008 4282 4a80 6708 0e98 2800 5380 <66f8> 0801 0001 6704 0e58 2800 0801 0000 6704 0e10 2800 241f 4e5e 4e75 4e75 4e56 [ 303.560000] Disabling lock debugging due to kernel taint
root@(none):/#
Also, it appears that the patch Michael posted in this thread would have prevented that oops because the "jne 1b" at __clear_user+0x22 would have been found in the exception table.
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Hi Finn,
thanks for (inadvertently) proving this bug in __clear_user() does exist!
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below? The previous bug reports might be considered somewhat contrived but this one's from 'real' user space code, and none too complex at that?
Cheers,
Michael
Quoted 147 lines · Finn Thain
Quoted 22 lines · Michael Schmitz
As mentioned by Finn Thain in his patch to improve
put_userexception handling on 040, a similar problem exists on 030 processors.A moves instruction that crosses a page boundary from a mapped page into an unmapped one will cause a mid-instruction bus error exception (frame format b), with the PC pointing (usually) two instructions past the faulting movesl instruction.
Our exception handling in
__generic_copy_to_useronly covers the instruction immediately following the faulting one. As a result,fixup_exceptioninsend_fault_sigdoes not detect this case, and causesend_fault_sigto oops.Extend the exception table to cover one additional instruction beyond the moves[lwb] instructions.
Tested on 68030 (Atari Falcon 030) with transfers beginning at one to six bytes offset from the end of a mapped page, followed by further bytes on an unmapped page (testcase derived from stress-ng sysbadaddr stressor by Finn Thain).
Quoted 6 lines
Tested on 68040 (Mac Quadra) and 68030 (Mac IIci) by Finn Thain.
A similar problem is present in
__clear_user(); modify the exception table for that function in the same way (tested by Finn Thain).Well, that
__clear_user()bug is no longer theoretical. I accidentally bumped into it when I sent a ^C to a shell script I wrote to test somemac_scsidriver patches...[ 0.000000] Linux version 6.10.0-mac-00011-gd1d490afeb2a (fthain@nippy) (m68k-unknown-linux-musl-gcc (Gentoo 13.2.1_p20240210 p14) 13.2.1 20240210, GNU ld (Gentoo 2.40 p7) 2.40.0) #39 Sat Aug 3 19:57:43 AEST 2024...
root@(none):/# bash scsi-test.sh /dev/sda5 bs=512k count=1 [ 130.090000] sd 0:0:0:0: [sda] tag#9 PDMA fixup: DRQ timeout [ 130.090000] sd 0:0:0:0: [sda] tag#9 switching to slow handshake [ 130.180000] sd 0:0:0:0: Power-on or device reset occurred 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 5.13732 s, 102 kB/s [ 136.030000] bash (57): drop_caches: 3 ^C root@(none):/# bash scsi-test.sh /dev/sdb5 bs=512k count=1 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.1273 s, 465 kB/s [ 160.440000] bash (61): drop_caches: 3 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.12763 s, 465 kB/s [ 171.920000] bash (61): drop_caches: 3bs=64k count=8
8+0 records in
8+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.03469 s, 507 kB/s[ 184.480000] bash (61): drop_caches: 3 8+0 records in 8+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.03221 s, 508 kB/s [ 196.340000] bash (61): drop_caches: 3bs=4k count=128
128+0 records in
128+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.13089 s, 464 kB/s[ 208.860000] bash (61): drop_caches: 3 128+0 records in 128+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.13089 s, 464 kB/s [ 220.850000] bash (61): drop_caches: 3bs=512 count=1k
1024+0 records in
1024+0 records out
524288 bytes (524 kB, 512 KiB) copied, 3.69678 s, 142 kB/s[ 236.020000] bash (61): drop_caches: 3 1024+0 records in 1024+0 records out 524288 bytes (524 kB, 512 KiB) copied, 3.68874 s, 142 kB/s [ 250.570000] bash (61): drop_caches: 3bs=512k count=1
1+0 records in
1+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.12338 s, 467 kB/s[ 263.120000] bash (61): drop_caches: 3 1+0 records in 1+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.12665 s, 465 kB/s [ 275.110000] bash (61): drop_caches: 3bs=64k count=8
8+0 records in
8+0 records out
524288 bytes (524 kB, 512 KiB) copied, 1.08861 s, 482 kB/s[ 287.610000] bash (61): drop_caches: 3 8+0 records in 8+0 records out 524288 bytes (524 kB, 512 KiB) copied, 1.03319 s, 507 kB/s [ 299.510000] bash (61): drop_caches: 3 ^C[ 303.190000] Unable to handle kernel access at virtual address d7d17fd7 [ 303.200000] Oops: 00000000 [ 303.210000] Modules linked in: [ 303.220000] PC: [<004301fe>] __clear_user+0x22/0x40 [ 303.240000] SR: 2000 SP: 4841d159 a2: 00b78590 [ 303.250000] d0: 000003ab d1: 00000000 d2: 00000000 d3: 8000c150 [ 303.260000] d4: 00002000 d5: 00000000 a0: 8000c154 a1: 009bdb3c [ 303.280000] Process cmp (pid: 92, task=7d95deea) [ 303.290000] Frame format=B ssw=0709 isc=0801 isb=0001 daddr=8000c150 dobuf=00000000 [ 303.310000] baddr=00430202 dibuf=8000c150 ver=f [ 303.320000] Stack from 0149de8c: [ 303.320000] 8001c2c4 0149deb8 00139904 8000c150 00000eb0 00983c00 00000000 00000003 [ 303.320000] 00000003 8000bf00 00983a00 0149df40 0013a1fa 00912480 8000bf00 00983c60 [ 303.320000] 00000003 00000012 00000000 fffffff8 00000006 d012f728 00000000 00983a5a [ 303.320000] 00000004 00983a00 000579d4 0052b7e8 009125a0 00912480 00983a5a 80000034 [ 303.320000] 000d1704 014994c0 80008864 80000000 00000000 80008864 80008864 00000000 [ 303.320000] 01497800 00000000 00000034 fffffff8 fffffff8 0149df7c 000e9296 00983a00 [ 303.380000] Call Trace: [<00139904>] elf_load+0x192/0x1da [ 303.390000] [<0013a1fa>] load_elf_binary+0x8ae/0xe46 [ 303.400000] [<000579d4>] try_module_get+0x0/0x48 [ 303.410000] [<000d1704>] kfree+0x0/0x160 [ 303.420000] [<000e9296>] bprm_execve+0x134/0x416 [ 303.430000] [<000e9daa>] copy_string_kernel+0x0/0x184 [ 303.440000] [<000e9fa4>] copy_strings+0x0/0x310 [ 303.460000] [<000ea3f6>] do_execveat_common+0x142/0x1d2 [ 303.480000] [<000eb128>] sys_execve+0x2a/0x36 [ 303.490000] [<0000275e>] syscall+0x8/0xc [ 303.510000] [<0008c010>] do_read_cache_folio+0xde/0x2c8 [ 303.520000] [ 303.520000] Code: 206e 0008 4282 4a80 6708 0e98 2800 5380 <66f8> 0801 0001 6704 0e58 2800 0801 0000 6704 0e10 2800 241f 4e5e 4e75 4e75 4e56 [ 303.560000] Disabling lock debugging due to kernel taintroot@(none):/#
Also, it appears that the patch Michael posted in this thread would have prevented that oops because the "jne 1b" at
__clear_user+0x22would have been found in the exception table.
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Hello Michael
Quoted 3 lines · Michael Schmitz
Quoted 3 lines
Well, that
__clear_user()bug is no longer theoretical. I accidentally bumped into it when I sent a ^C to a shell script I wrote to test somemac_scsidriver patches...
...
Quoted 3 lines
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below?
Apparently the series is waiting for some testing on a Coldfire system with MMU.
Quoted 3 lines
The previous bug reports might be considered somewhat contrived but this one's from 'real' user space code, and none too complex at that?
Right. That code was as follows. There's nothing here aimed at arch/m68k/lib/uaccess.c in particular, just IO to a block device and a tmpfs filesystem.
#!/bin/bash
set -e -u
filename=$1
rand=/tmp/rand_test_data
zero=/tmp/zero_test_data
len=512K
dd if=/dev/urandom of=$rand bs=$len count=1 2>/dev/null dd if=/dev/zero of=$zero bs=$len count=1 2>/dev/null
write() {
dd $args if=$orig of=$filename # oflag=direct
}
compare() {
echo 3 > /proc/sys/vm/drop_caches
if ! cmp -n $len $orig $filename ; then
diff -u <(hexdump -C < $orig) <(hexdump -C < $filename)
fi
}
while true; do
for args in "bs=512k count=1" "bs=64k count=8" "bs=4k count=128" "bs=512 count=1k" ; do
echo $args
orig=$rand ; write ; compare
orig=$zero ; write ; compare
sync
echo
done
done
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Hi Michael, Finn,
Quoted 9 lines · Finn Thain
Hello Michael
Quoted 3 lines · Michael Schmitz
Quoted 3 lines
Well, that
__clear_user()bug is no longer theoretical. I accidentally bumped into it when I sent a ^C to a shell script I wrote to test somemac_scsidriver patches...
Quoted 8 lines
...
Quoted 3 lines
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below?
Apparently the series is waiting for some testing on a Coldfire system with MMU.
Ok, I am in a state that I can do that now (I managed to fix my M5475EVB board). If I test the v4 versions of this patch set that should do the job?
Regards
Greg
Quoted 42 lines
Quoted 3 lines
The previous bug reports might be considered somewhat contrived but this one's from 'real' user space code, and none too complex at that?
Right. That code was as follows. There's nothing here aimed at
arch/m68k/lib/uaccess.cin particular, just IO to a block device and a tmpfs filesystem.#!/bin/bash
set -e -u
filename=$1
rand=/tmp/rand_test_data
zero=/tmp/zero_test_data
len=512Kdd if=/dev/urandom of=$rand bs=$len count=1 2>/dev/null dd if=/dev/zero of=$zero bs=$len count=1 2>/dev/null
write(){
dd $args if=$orig of=$filename # oflag=direct
}
compare(){
echo 3 >/proc/sys/vm/drop_caches
if ! cmp -n $len $orig $filename ; then
diff -u <(hexdump -C < $orig) <(hexdump -C < $filename)
fi
}while true; do
for args in "bs=512k count=1" "bs=64k count=8" "bs=4k count=128" "bs=512 count=1k" ; do
echo $args
orig=$rand ; write ; compare
orig=$zero ; write ; compare
sync
echo
done
done
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Quoted 11 lines · Greg Ungerer
Quoted 7 lines · Finn Thain
Quoted 4 lines
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below?
Apparently the series is waiting for some testing on a Coldfire system with MMU.
Ok, I am in a state that I can do that now (I managed to fix my M5475EVB board).
Thanks, Greg.
Quoted 2 lines
If I test the v4 versions of this patch set that should do the job?
There are 3 patches that need some more testing, one from me and two from Michael:
[PATCH] m68k: Handle put_user() faults more carefully
[PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully [PATCH v4 2/2] m68k: improve __constant_copy_to_user_asm() fault handling
They were posted in two threads:
https://lore.kernel.org/linux-m68k/1ed9c4c753395510c1a8df9c35e2ad4c31c90f95.1714265926.git.fthain@linux-m68k.org/T/
https://lore.kernel.org/linux-m68k/CAMuHMdVrOnOQwCxk42YCjEkbfL-YDSvpf_xTaouv9hUs2bO+qg@mail.gmail.com/T/
On 680x0, one of the bugs was brought to light with 'stress-ng --sysbadaddr -1'. The others required special test programs.
I've no idea whether Coldfire silicon is susceptable at all, and if it is, whether the same reproducers would work. But that's neither here nor there from a regression testing standpoint.
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
In reply to Greg UngererHi Greg,
Am 08.08.2024 um 18:05 schrieb Greg Ungerer:
Quoted 18 lines
Hi Michael, Finn,
Quoted 15 lines · Finn Thain
Hello Michael
Quoted 3 lines · Michael Schmitz
Quoted 3 lines
Well, that
__clear_user()bug is no longer theoretical. I accidentally bumped into it when I sent a ^C to a shell script I wrote to test somemac_scsidriver patches......
Quoted 3 lines
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below?
Quoted 6 lines
Quoted 2 lines
Apparently the series is waiting for some testing on a Coldfire system with MMU.
Ok, I am in a state that I can do that now (I managed to fix my M5475EVB board).
If I test the v4 versions of this patch set that should do the job?
Yes, v4 is the one to test.
I'll see that I can dig out my test code to reproduce the bug.
Thanks,
Michael
Quoted 49 lines
Regards
GregQuoted 43 lines
Quoted 3 lines
The previous bug reports might be considered somewhat contrived but this one's from 'real' user space code, and none too complex at that?
Right. That code was as follows. There's nothing here aimed at
arch/m68k/lib/uaccess.cin particular, just IO to a block device and a tmpfs filesystem.#!/bin/bash
set -e -u
filename=$1
rand=/tmp/rand_test_data
zero=/tmp/zero_test_data
len=512Kdd if=/dev/urandom of=$rand bs=$len count=1 2>/dev/null dd if=/dev/zero of=$zero bs=$len count=1 2>/dev/null
write(){
dd $args if=$orig of=$filename # oflag=direct
}
compare(){
echo 3 >/proc/sys/vm/drop_caches
if ! cmp -n $len $orig $filename ; then
diff -u <(hexdump -C < $orig) <(hexdump -C < $filename)
fi
}while true; do
for args in "bs=512k count=1" "bs=64k count=8" "bs=4kcount=128" "bs=512 count=1k" ; do
echo $args
orig=$rand ; write ; compare
orig=$zero ; write ; compare
sync
echo
done
done
Greg Ungerer
Tested-byRe: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
In reply to Finn ThainHi Finn,
Quoted 28 lines · Finn Thain
Quoted 9 lines · Greg Ungerer
Quoted 5 lines · Finn Thain
Quoted 2 lines
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below?
Apparently the series is waiting for some testing on a Coldfire system with MMU.
Ok, I am in a state that I can do that now (I managed to fix my M5475EVB board).
Thanks, Greg.
Quoted 2 lines
If I test the v4 versions of this patch set that should do the job?
There are 3 patches that need some more testing, one from me and two from Michael:
[PATCH] m68k: Handle
put_user()faults more carefully
[PATCH v4 1/2] m68k: Handle__generic_copy_to_userfaults more carefully [PATCH v4 2/2] m68k: improve__constant_copy_to_user_asm()fault handlingThey were posted in two threads:
https://lore.kernel.org/linux-m68k/1ed9c4c753395510c1a8df9c35e2ad4c31c90f95.1714265926.git.fthain@linux-m68k.org/T/
https://lore.kernel.org/linux-m68k/CAMuHMdVrOnOQwCxk42YCjEkbfL-YDSvpf_xTaouv9hUs2bO+qg@mail.gmail.com/T/
Quoted 6 lines
On 680x0, one of the bugs was brought to light with 'stress-ng --sysbadaddr -1'. The others required special test programs.
I've no idea whether Coldfire silicon is susceptable at all, and if it is, whether the same reproducers would work. But that's neither here nor there from a regression testing standpoint.
Ok, thanks for the links. I have applied and tested those, no obvious regressions. So for all of these patches:
I tried out the "stress-ng --sysbadaddr -1" test, and that didn't go so well for me:
# stress-ng --sysbadaddr -1
stress-ng: info: [37] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor
stress-ng: info: [37] dispatching hogs: 1 sysbadaddr
*** ILLEGAL INSTRUCTION *** FORMAT=4
Current process id is 39
BAD KERNEL TRAP: 00000000
Modules linked in:
PC: [<00000000>] 0x0
SR: 2004 SP: 6504e563 a2: 008ee380
d0: 000000f7 d1: 00000000 d2: 00000000 d3: 00000000
d4: 00a87b80 d5: bfbf3814 a0: 00000000 a1: bfbf3814
Process stress-ng-sysba (pid: 39, task=4dbb2ec5)
Frame format=4 eff addr=480a2004 pc=0002b154
Stack from 00adff20:
00ade000 00000000 00000000 000000f7 00000000 00000004 00a87b80 00000000
00000000 00000000 00000000 008ee380 0002ab5c 00000100 00000122 fffffff6
bfbf376c 0002b29e 000000f7 bfbf3814 00000000 00000000 00ade000 0002b222
00ae0800 80118988 00000000 00000005 bfbf37a0 00000005 bfbf3814 00adffcc
00023d2c 00adffcc 00000000 00000000 00000000 00000000 000000f7 00000000
80118b46 00021850 00024b00 000000f7 bfbf3814 00000000 00000000 bfbf3814
Call Trace: [<0002ab5c>] child_wait_callback+0x0/0x24
[<0002b29e>] sys_wait4+0x7c/0x8e
[<0002b222>] sys_wait4+0x0/0x8e
[<00023d2c>] buserr_c+0xb0/0x152
[<00021850>] buserr+0x28/0x30
[<00024b00>] system_call+0x54/0xa8But that is the same with and without these patches.
Regards
Greg
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Hi Greg,
tanks for testing!
Quoted 74 lines · Greg Ungerer
Hi Finn,
Quoted 30 lines · Finn Thain
Quoted 9 lines · Greg Ungerer
Quoted 5 lines · Finn Thain
Quoted 2 lines
I'm afraid I've lost track of where we're at with this patch series. Does it need more work, or more bug reports such as the one below?
Apparently the series is waiting for some testing on a Coldfire system with MMU.
Ok, I am in a state that I can do that now (I managed to fix my M5475EVB board).
Thanks, Greg.
Quoted 2 lines
If I test the v4 versions of this patch set that should do the job?
There are 3 patches that need some more testing, one from me and two from Michael:
[PATCH] m68k: Handle
put_user()faults more carefully
[PATCH v4 1/2] m68k: Handle__generic_copy_to_userfaults more carefully [PATCH v4 2/2] m68k: improve__constant_copy_to_user_asm()fault handlingThey were posted in two threads:
Quoted 6 lines
On 680x0, one of the bugs was brought to light with 'stress-ng --sysbadaddr -1'. The others required special test programs.
I've no idea whether Coldfire silicon is susceptable at all, and if it is, whether the same reproducers would work. But that's neither here nor there from a regression testing standpoint.
Ok, thanks for the links. I have applied and tested those, no obvious regressions. So for all of these patches:
Tested-by Greg Ungerer <[email protected]>I tried out the "stress-ng --sysbadaddr -1" test, and that didn't go so well for me:
# stress-ng --sysbadaddr -1 stress-ng: info: [37] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor stress-ng: info: [37] dispatching hogs: 1 sysbadaddr *** ILLEGAL INSTRUCTION *** FORMAT=4 Current process id is 39 BAD KERNEL TRAP: 00000000 Modules linked in: PC: [<00000000>] 0x0 SR: 2004 SP: 6504e563 a2: 008ee380 d0: 000000f7 d1: 00000000 d2: 00000000 d3: 00000000 d4: 00a87b80 d5: bfbf3814 a0: 00000000 a1: bfbf3814 Process stress-ng-sysba (pid: 39, task=4dbb2ec5) Frame format=4 eff addr=480a2004 pc=0002b154 Stack from 00adff20: 00ade000 00000000 00000000 000000f7 00000000 00000004 00a87b80 00000000 00000000 00000000 00000000 008ee380 0002ab5c 00000100 00000122 fffffff6 bfbf376c 0002b29e 000000f7 bfbf3814 00000000 00000000 00ade000 0002b222 00ae0800 80118988 00000000 00000005 bfbf37a0 00000005 bfbf3814 00adffcc 00023d2c 00adffcc 00000000 00000000 00000000 00000000 000000f7 00000000 80118b46 00021850 00024b00 000000f7 bfbf3814 00000000 00000000 bfbf3814 Call Trace: [<0002ab5c>] child_wait_callback+0x0/0x24 [<0002b29e>] sys_wait4+0x7c/0x8e [<0002b222>] sys_wait4+0x0/0x8e [<00023d2c>] buserr_c+0xb0/0x152 [<00021850>] buserr+0x28/0x30 [<00024b00>] system_call+0x54/0xa8
Quoted 1 line
But that is the same with and without these patches.
I wonder if recent signal handling changes (e.g. commit 0d4276cfbe6fd4c4a21acdee803b05a3a6192082) have rare unexpected side effects on Coldfire here ... OTOH, signal handling as such works just fine, right?
Cheers,
Michael
Quoted 4 lines
Regards
Greg
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
In reply to Greg UngererQuoted 7 lines · Greg Ungerer
Ok, thanks for the links. I have applied and tested those, no obvious regressions.
So for all of these patches:Tested-by Greg Ungerer <[email protected]>
Thanks for testing!
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
In reply to Michael SchmitzQuoted 38 lines · Michael Schmitz
Quoted 38 lines
I tried out the "stress-ng --sysbadaddr -1" test, and that didn't go so well for me:
# stress-ng --sysbadaddr -1 stress-ng: info: [37] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor stress-ng: info: [37] dispatching hogs: 1 sysbadaddr *** ILLEGAL INSTRUCTION *** FORMAT=4 Current process id is 39 BAD KERNEL TRAP: 00000000 Modules linked in: PC: [<00000000>] 0x0 SR: 2004 SP: 6504e563 a2: 008ee380 d0: 000000f7 d1: 00000000 d2: 00000000 d3: 00000000 d4: 00a87b80 d5: bfbf3814 a0: 00000000 a1: bfbf3814 Process stress-ng-sysba (pid: 39, task=4dbb2ec5) Frame format=4 eff addr=480a2004 pc=0002b154 Stack from 00adff20: 00ade000 00000000 00000000 000000f7 00000000 00000004 00a87b80 00000000 00000000 00000000 00000000 008ee380 0002ab5c 00000100 00000122 fffffff6 bfbf376c 0002b29e 000000f7 bfbf3814 00000000 00000000 00ade000 0002b222 00ae0800 80118988 00000000 00000005 bfbf37a0 00000005 bfbf3814 00adffcc 00023d2c 00adffcc 00000000 00000000 00000000 00000000 000000f7 00000000 80118b46 00021850 00024b00 000000f7 bfbf3814 00000000 00000000 bfbf3814 Call Trace: [<0002ab5c>] child_wait_callback+0x0/0x24 [<0002b29e>] sys_wait4+0x7c/0x8e [<0002b222>] sys_wait4+0x0/0x8e [<00023d2c>] buserr_c+0xb0/0x152 [<00021850>] buserr+0x28/0x30 [<00024b00>] system_call+0x54/0xa8
Quoted 6 lines
Quoted 1 line
But that is the same with and without these patches.
I wonder if recent signal handling changes (e.g. commit
0d4276cfbe6fd4c4a21acdee803b05a3a6192082) have rare unexpected side effects on Coldfire here ... OTOH, signal handling as such works just fine, right?
That would be commit b845b574f86d ("m68k: Move signal frame following exception on 68020/030") on mainline. If it caused a regression, that would have first appeared in v6.4. I can't imagine how that commit could affect Coldfire but that's no reason not to test an older kernel. FWIW, my hunch is that the other stressors which call wait4() will probably crash too (regardless of kernel version).
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Hi Finn,
Am 09.08.2024 um 15:34 schrieb Finn Thain:
Quoted 48 lines · Michael Schmitz
Quoted 39 lines
Quoted 39 lines
I tried out the "stress-ng --sysbadaddr -1" test, and that didn't go so well for me:
# stress-ng --sysbadaddr -1 stress-ng: info: [37] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor stress-ng: info: [37] dispatching hogs: 1 sysbadaddr *** ILLEGAL INSTRUCTION *** FORMAT=4 Current process id is 39 BAD KERNEL TRAP: 00000000 Modules linked in: PC: [<00000000>] 0x0 SR: 2004 SP: 6504e563 a2: 008ee380 d0: 000000f7 d1: 00000000 d2: 00000000 d3: 00000000 d4: 00a87b80 d5: bfbf3814 a0: 00000000 a1: bfbf3814 Process stress-ng-sysba (pid: 39, task=4dbb2ec5) Frame format=4 eff addr=480a2004 pc=0002b154 Stack from 00adff20: 00ade000 00000000 00000000 000000f7 00000000 00000004 00a87b80 00000000 00000000 00000000 00000000 008ee380 0002ab5c 00000100 00000122 fffffff6 bfbf376c 0002b29e 000000f7 bfbf3814 00000000 00000000 00ade000 0002b222 00ae0800 80118988 00000000 00000005 bfbf37a0 00000005 bfbf3814 00adffcc 00023d2c 00adffcc 00000000 00000000 00000000 00000000 000000f7 00000000 80118b46 00021850 00024b00 000000f7 bfbf3814 00000000 00000000 bfbf3814 Call Trace: [<0002ab5c>] child_wait_callback+0x0/0x24 [<0002b29e>] sys_wait4+0x7c/0x8e [<0002b222>] sys_wait4+0x0/0x8e [<00023d2c>] buserr_c+0xb0/0x152 [<00021850>] buserr+0x28/0x30 [<00024b00>] system_call+0x54/0xa8Quoted 6 lines
Quoted 1 line
But that is the same with and without these patches.
I wonder if recent signal handling changes (e.g. commit
0d4276cfbe6fd4c4a21acdee803b05a3a6192082) have rare unexpected side effects on Coldfire here ... OTOH, signal handling as such works just fine, right?
Quoted 4 lines
That would be commit
b845b574f86d("m68k: Move signal frame following exception on 68020/030") on mainline. If it caused a regression, that would have first appeared in v6.4. I can't imagine how that commit could affect Coldfire but that's no reason not to test an older kernel.
Yes, testing older kernels is certainly the fastest way to rule out involvement of that commit.
On a closer look, there is no possible way in that this commit can be responsible for the bug unless Coldfire does use frame format B for access errors. I don't think that's likely?
Quoted 2 lines
FWIW, my hunch is that the other stressors which call
wait4()will probably crash too (regardless of kernel version).
Quite likely ...
Cheers,
Michael
Re: [PATCH v4 1/2] m68k: Handle __generic_copy_to_user faults more carefully
Hi Michael, Finn,
Quoted 51 lines · Michael Schmitz
Hi Finn,
Am 09.08.2024 um 15:34 schrieb Finn Thain:
Quoted 48 lines · Michael Schmitz
Quoted 45 lines
Quoted 39 lines
I tried out the "stress-ng --sysbadaddr -1" test, and that didn't go so well for me:
# stress-ng --sysbadaddr -1 stress-ng: info: [37] defaulting to a 86400 second (1 day, 0.00 secs) run per stressor stress-ng: info: [37] dispatching hogs: 1 sysbadaddr *** ILLEGAL INSTRUCTION *** FORMAT=4 Current process id is 39 BAD KERNEL TRAP: 00000000 Modules linked in: PC: [<00000000>] 0x0 SR: 2004 SP: 6504e563 a2: 008ee380 d0: 000000f7 d1: 00000000 d2: 00000000 d3: 00000000 d4: 00a87b80 d5: bfbf3814 a0: 00000000 a1: bfbf3814 Process stress-ng-sysba (pid: 39, task=4dbb2ec5) Frame format=4 eff addr=480a2004 pc=0002b154 Stack from 00adff20: 00ade000 00000000 00000000 000000f7 00000000 00000004 00a87b80 00000000 00000000 00000000 00000000 008ee380 0002ab5c 00000100 00000122 fffffff6 bfbf376c 0002b29e 000000f7 bfbf3814 00000000 00000000 00ade000 0002b222 00ae0800 80118988 00000000 00000005 bfbf37a0 00000005 bfbf3814 00adffcc 00023d2c 00adffcc 00000000 00000000 00000000 00000000 000000f7 00000000 80118b46 00021850 00024b00 000000f7 bfbf3814 00000000 00000000 bfbf3814 Call Trace: [<0002ab5c>] child_wait_callback+0x0/0x24 [<0002b29e>] sys_wait4+0x7c/0x8e [<0002b222>] sys_wait4+0x0/0x8e [<00023d2c>] buserr_c+0xb0/0x152 [<00021850>] buserr+0x28/0x30 [<00024b00>] system_call+0x54/0xa8Quoted 1 line
But that is the same with and without these patches.
I wonder if recent signal handling changes (e.g. commit
0d4276cfbe6fd4c4a21acdee803b05a3a6192082) have rare unexpected side effects on Coldfire here ... OTOH, signal handling as such works just fine, right?
Quoted 6 lines
Quoted 4 lines
That would be commit
b845b574f86d("m68k: Move signal frame following exception on 68020/030") on mainline. If it caused a regression, that would have first appeared in v6.4. I can't imagine how that commit could affect Coldfire but that's no reason not to test an older kernel.Yes, testing older kernels is certainly the fastest way to rule out involvement of that commit.
I will go back a few revisions and see if I can shed some light on it.
Regards
Greg
Quoted 13 lines
On a closer look, there is no possible way in that this commit can be responsible for the bug unless Coldfire does use frame format B for access errors. I don't think that's likely?
Quoted 2 lines
FWIW, my hunch is that the other stressors which call
wait4()will probably crash too (regardless of kernel version).Quite likely ...
Cheers,
Michael