moxa serial driver testing
Subsystems (2)
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
A message that is not in the archive (sent off-list or not yet received): [email protected]
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Please enable
* Lock debugging: prove locking correctness
* Lock dependency engine debugging
at Kernel hacking->Kernel debugging
Quoted 4 lines
2. Another, less important problem, is that if a port is open by some application and I rmmod the mxser (or mxser-new) module, kernel oopses and gets unstable ("xjterm" is the process that keeps the port open):
I'll look into that.
Quoted 44 lines
Couldn't unregister MOXA Smartio/Industio family serial driver
BUG: unable to handle kernel paging request at virtual address f0fe60be printing eip: f0fe60be *pde = 3fc04067 *pte = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod
sr_modsbp2 ieee1394ide_genericide_diske1000snd_hda_intelsnd_hda_codecpsmousei2c_i801mousedevsnd_pcm_osssnd_mixer_ossserio_rawi2c_coretsdev 8250_pnp pcspkrparport_pcevdevsnd_pcm8250_pci parportsnd_timerrtc floppy snd soundcoresnd_page_allocext3 jbd mbcachesd_modusb_storageusbhidide_cdcdromuhci_hcdata_piixlibata usbcore piixscsi_modgenericide_coreskge thermal processor fanCPU: 0 EIP: 0060:[<f0fe60be>] Tainted: P VLI EFLAGS: 00210246 (2.6.17-2-686 #1)EIP is at
0xf0fe60be
eax:ebc8c800ebx:ebc8c800ecx: 00000000 edx:efe4e400
esi: 00000000 edi:e2877a80ebp:ebc8c80cesp:d924dbf8
ds: 007b es: 007b ss: 0068
Process xjterm (pid: 11884, threadinfo=d924c000 task=e6835a90)Stack: b01f550c 00123fc3 ebc8c800 b01f5418 e2877a80 b01f170a 00000000 00000000 00000020 e2877a80 d924de98 d924dea8 b0162749 d924dfb0 080e8de0 080e8df8 d924de98 d924de98 00000000 00000003 d924deb0 00000002 00000000 d924de98 Call Trace: <b01f550c> normal_poll+0xf4/0x119 <b01f5418> normal_poll+0x0/0x119 <b01f170a> tty_poll+0x48/0x60 <b0162749> do_sys_poll+0x1af/0x342 <b01631fd> __pollwait+0x0/0xb2 <b0116f2d> default_wake_function+0x0/0xc <b0116f2d> default_wake_function+0x0/0xc <b0116f2d> default_wake_function+0x0/0xc <b0116f2d> default_wake_function+0x0/0xc <f182310a> _nv002668rm+0x26/0x2c [nvidia] <b013f030> __alloc_pages+0x4e/0x267 <b0116b0b> activate_task+0x5a/0xa0 <b01160df> __activate_task+0x17/0x1d <b0116f23> try_to_wake_up+0x319/0x323 <b0116b0b> activate_task+0x5a/0xa0 <b01160df> __activate_task+0x17/0x1d <b0116f23> try_to_wake_up+0x319/0x323 <b0115f0a> __wake_up_common+0x2f/0x53 <b0117cb2> __wake_up+0x2a/0x3d <b0217309> sock_def_readable+0x31/0x5b <b0267d05> unix_stream_sendmsg+0x200/0x2c4 <b0213924> do_sock_write+0xa3/0xaa <b0213e98> sock_aio_write+0x53/0x61 <b0152afc> do_sync_write+0xb8/0xf3 <b012c2a7> autoremove_wake_function+0x0/0x2d <b026744e> unix_ioctl+0x8b/0x93 <b021409a> sock_ioctl+0x193/0x1b5 <b0213f07> sock_ioctl+0x0/0x1b5 <b0161e08> do_ioctl+0x1c/0x5d <b0162093> vfs_ioctl+0x24a/0x25c <b0162920> sys_poll+0x44/0x47 <b0102ae7> sysenter_past_esp+0x54/0x75 Code: Bad EIP value.
Quoted 5 lines
EIP: [<
f0fe60be>]0xf0fe60beSS:ESP 0068:d924dbf8Note: I'm running Debian kernel 2.6.17.2 but I've seen the same problem on 2.6.16.x. Didn't try more recent kernels though. CPU is hyper-threaded Pentium-4.
thanks for the report,
Signature · 4 lines
http://www.fi.muni.cz/~xslaby/ Jiri Slaby faculty of informatics, masaryk university, brno, cz e-mail: jirislaby gmail com, gpg pubkey fingerprint: B674 9967 0407 CE62 ACC8 22A0 32CC 55C3 39D4 7A7E
Quoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Please enable
BTW. Does sysrq-keys work in such situation for you? I'll appreciate if you was able to grab sysrq-t output, when it happens.
thanks,
Signature · 4 lines
http://www.fi.muni.cz/~xslaby/ Jiri Slaby faculty of informatics, masaryk university, brno, cz e-mail: jirislaby gmail com, gpg pubkey fingerprint: B674 9967 0407 CE62 ACC8 22A0 32CC 55C3 39D4 7A7E
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
---
drivers/char/mxser_new.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-)
drivers/char/mxser_new.c +4 -2
@@ -2268,6 +2268,8 @@ static irqreturn_t mxser_interrupt(int irq, void *dev_id)if (bits & irqbits)continue ; port = &brd->ports[i]; +if (!(port->flags &ASYNC_INITIALIZED )) +continue ; int_cnt =0 ; do {
@@ -2320,9 +2322,9 @@ static irqreturn_t mxser_interrupt(int irq, void *dev_id)if (status &UART_LSR_THRE )mxser_transmit_chars (port); } - }while (int_cnt++ <MXSER_ISR_PASS_LIMIT ); + }while (int_cnt++ <256 ); } -if (pass_counter++ >MXSER_ISR_PASS_LIMIT ) +if (pass_counter++ >64 )break ;/* Prevent infinite loops */ }
Jiri,
Quoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
Thanks for looking into it. I'll be able to get to the box with moxa installed on Monday and will try the patch.
As for SysRq, I'm afraid it didn't work though I'm not 100% sure. I'll check that as well.
-- Sergei.
Quoted 30 lines
---
drivers/char/mxser_new.c | 6 ++++-- 1 files changed, 4 insertions(+), 2 deletions(-)drivers/char/mxser_new.c +4
-2@@ -2268,6 +2268,8 @@ static irqreturn_t mxser_interrupt(int irq, void *dev_id)if (bits & irqbits)continue ; port = &brd->ports[i]; +if (!(port->flags &ASYNC_INITIALIZED )) +continue ; int_cnt =0 ; do {@@ -2320,9 +2322,9 @@ static irqreturn_t mxser_interrupt(int irq, void *dev_id)if (status &UART_LSR_THRE )mxser_transmit_chars (port); } - }while (int_cnt++ <MXSER_ISR_PASS_LIMIT ); + }while (int_cnt++ <256 ); } -if (pass_counter++ >MXSER_ISR_PASS_LIMIT ) +if (pass_counter++ >64 )break ;/* Prevent infinite loops */ }
Quoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
I'm preparing to test it. However, it seems that my version of mxser_new.c got out of sync with your one:
osv@osv mxser_new$ patch -p3 < lock.patch patching file mxser_new.c
Hunk #1 succeeded at 1900 (offset -368 lines). Hunk #2 succeeded at 1968 (offset -354 lines). osv@osv mxser_new$ patch -p3 < rmmod.patch patching file mxser.c
Hunk #1 succeeded at 711 with fuzz 2 (offset -6 lines). patching file mxser_new.c
Hunk #1 succeeded at 705 with fuzz 2 (offset -1985 lines). osv@osv mxser_new$
I'll try this one anyway, but could you please either tell me where to get the version you are using, or just send it to me by mail?
BTW, when system hangs, SysRq magic doesn't work anymore.
-- Sergei.
Quoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
Something did change. Now it becomes rather difficult to get the box to hang, though not impossible. Another thing that changed is that now I can see [parts of] oopses on screen. I've got two pictures of the screen with different oopses. If you need them, let me know and I'll send them to you in a separate mail not to pollute lkml with JPEGs.
One of these oopses happened on kernel with lock debugging enabled, but I can't see anything relevant in dmesg log after resetting the box.
SysRq magic still doesn't work after hang.
As for the problem with module unloading when port is open, your another patch does fix it indeed.
-- Sergei.
Quoted 8 lines · Sergei Organov
Quoted 6 lines · Jiri Slaby
Quoted 5 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
Quoted 8 lines
Quoted 4 lines
Quoted 3 lines
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
Something did change. Now it becomes rather difficult to get the box to hang, though not impossible. Another thing that changed is that now I can see [parts of] oopses on screen. I've got two pictures of the screen
Positive.
Quoted 2 lines
with different oopses. If you need them, let me know and I'll send them to you in a separate mail not to pollute lkml with JPEGs.
These are those you've posted in another post?
Quoted 2 lines
As for the problem with module unloading when port is open, your another patch does fix it indeed.
at least some good news, thanks,
Signature · 4 lines
http://www.fi.muni.cz/~xslaby/ Jiri Slaby faculty of informatics, masaryk university, brno, cz e-mail: jirislaby gmail com, gpg pubkey fingerprint: B674 9967 0407 CE62 ACC8 22A0 32CC 55C3 39D4 7A7E
Quoted 18 lines · Jiri Slaby
Quoted 8 lines · Sergei Organov
Quoted 6 lines · Jiri Slaby
Quoted 5 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
Quoted 8 lines
Quoted 4 lines
Quoted 3 lines
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
Something did change. Now it becomes rather difficult to get the box to hang, though not impossible. Another thing that changed is that now I can see [parts of] oopses on screen. I've got two pictures of the screen
Quoted 6 lines
Positive.
Quoted 2 lines
with different oopses. If you need them, let me know and I'll send them to you in a separate mail not to pollute lkml with JPEGs.
These are those you've posted in another post?
Not exactly. The oopses seem to vary from run to run, so I can't get exactly the same every time. Though I kept those JPEGs, if you are still interested after looking at those oopses that I've collected through serial console and sent.
-- Sergei.
Re: moxa serial driver testing (oopses)
In reply to Jiri SlabyQuoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
In addition to my previous answer, fortunately I was able to log oopses to a serial console, so below are two of them.
They seem to appear from in a few seconds to in a few minutes after I run:
$ while true; do cat /dev/ttyM7 > /dev/null; done
when /dev/ttyM7 is setup so that 'cat' immediately returns (due to zero timeout).
Note that unpatched version always hangs in less than a second.
First oops is taken from the kernel with lock debugging enabled:
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000080 printing eip: f8fa5136 *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer snd evdev mxser_new tsdev soundcore psmouse snd_page_alloc serio_raw floppy i2c_i801 parport_pc parport pcspkr i2c_core rtc ext3 jbd mbcache usb_storage sd_mod usbhid ide_cd cdrom ata_piix libata uhci_hcd scsi_mod piix generic usbcore ide_core skge thermal processor fan CPU: 0 EIP: 0060:[<f8fa5136>] Not tainted VLI EFLAGS: 00010046 (2.6.18 #1)
EIP is at mxser_stoprx+0x5a/0x7e [mxser_new]
eax: 00000000 ebx: f7a55000 ecx: f8fabc1c edx: f76f6340
esi: c1e4651b edi: eb466fff ebp: f3af3de8 esp: f3af3de4
ds: 007b es: 007b ss: 0068
Process cc1plus (pid: 9540, ti=f3af2000 task=f4012ab0 task.ti=f3af2000)
Stack: f7a55000 f3af3df0 f8fa5162 f3af3ec0 c020aedb 000000ff c1e4641c f7a554dc
00000002 00000008 00000000 00000008 f4012ff0 00000001 0000017d 00000000
0000017d f4012ff0 00000001 0000017d 00000000 f4012ff0 0000017d 00000000
Call Trace:
[<c0103e2f>] show_stack_log_lvl+0x8c/0x97
[<c0103fa6>] show_registers+0x130/0x19d
[<c0104194>] die+0x181/0x287
[<c0115b92>] do_page_fault+0x3ca/0x49e
[<c01039cd>] error_code+0x39/0x40
[<f8fa5162>] mxser_throttle+0x8/0xa [mxser_new]
[<c020aedb>] n_tty_receive_buf+0xdce/0xdf0
[<c02062fc>] flush_to_ldisc+0x112/0x149
[<c0206370>] tty_flip_buffer_push+0x3d/0x53
[<f8fa63c4>] mxser_receive_chars+0x23c/0x244 [mxser_new]
[<f8fa6e48>] mxser_interrupt+0x14c/0x1cb [mxser_new]
[<c014563a>] handle_IRQ_event+0x20/0x4d
[<c01456fb>] __do_IRQ+0x94/0xef
[<c0105342>] do_IRQ+0x4e/0x60
[<c0103835>] common_interrupt+0x25/0x2c
Code: 83 e0 ee 89 41 3c 25 ee 00 00 00 42 eb 19 0f b6 42 1a 8b 51 08 89 41 38 31 c0 42 ee 89 d8 83 c8 02 89 41 3c 0f b6 c0 ee 8b 41 04 <8b> 80 80 00 00 00 83 78 08 00 79 15 8b 41 40 8b 51 08 83 e0 fd
EIP: [<f8fa5136>] mxser_stoprx+0x5a/0x7e [mxser_new] SS:ESP 0068:f3af3de4
<0>Kernel panic - not syncing: Fatal exception in interrupt
BUG: warning at arch/i386/kernel/smp.c:547/smp_call_function()
[<c0103e73>] show_trace+0xd/0x10
[<c010444c>] dump_stack+0x19/0x1b
[<c010f600>] smp_call_function+0x55/0xfd
[<c010f6be>] smp_send_stop+0x16/0x2a
[<c011d885>] panic+0x4d/0xec
[<c0104265>] die+0x252/0x287
[<c0115b92>] do_page_fault+0x3ca/0x49e
[<c01039cd>] error_code+0x39/0x40
[<f8fa5162>] mxser_throttle+0x8/0xa [mxser_new]
[<c020aedb>] n_tty_receive_buf+0xdce/0xdf0
[<c02062fc>] flush_to_ldisc+0x112/0x149
[<c0206370>] tty_flip_buffer_push+0x3d/0x53
[<f8fa63c4>] mxser_receive_chars+0x23c/0x244 [mxser_new]
[<f8fa6e48>] mxser_interrupt+0x14c/0x1cb [mxser_new]
[<c014563a>] handle_IRQ_event+0x20/0x4d
[<c01456fb>] __do_IRQ+0x94/0xef
[<c0105342>] do_IRQ+0x4e/0x60
[<c0103835>] common_interrupt+0x25/0x2cAnother oops is form kernel with lock debugging disabled:
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000068 printing eip: f8f0911f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm i2c_i801 psmouse snd_timer rtc parport_pc serio_raw i2c_core snd soundcore snd_page_alloc evdev pcspkr floppy parport tsdev mxser_new ext3 jbd mbcache usb_storage usbhid sd_mod ide_cd cdrom uhci_hcd ata_piix usbcore piix skge libata scsi_mod generic ide_core thermal processor fan CPU: 0 EIP: 0060:[<f8f0911f>] Tainted: P VLI EFLAGS: 00010246 (2.6.18-3-686 #1)
EIP is at mxser_stoprx+0x54/0x74 [mxser_new]
eax: 00000000 ebx: 00000001 ecx: f8f0f79c edx: f643a4c0
esi: efe6d51b edi: ee75dfff ebp: f6e2e000 esp: ec901e20
ds: 007b es: 007b ss: 0068
Process cc1plus (pid: 17764, ti=ec900000 task=dff41000 task.ti=ec900000)
Stack: 00000001 c01fe4af 000000ff efe6d41c f6e2e404 c02d6c68 00000000 00000001
ec901e64 c011669e 00000000 00000000 00000003 00000086 f6e2e00c 00000286
00000001 00000086 c01f97c9 00000000 f524bc00 00000286 f8f0a335 ec901ec8
Call Trace:
[<c01fe4af>] n_tty_receive_buf+0xcd6/0xcf9
[<c011669e>] __wake_up+0x2a/0x3d
[<c01f97c9>] tty_ldisc_deref+0x50/0x5f
[<f8f0a335>] mxser_receive_chars+0x241/0x249 [mxser_new]
[<f8f09678>] mxser_transmit_chars+0x14/0x164 [mxser_new]
[<f8f0adab>] mxser_interrupt+0x190/0x1e6 [mxser_new]
[<c0116412>] __activate_task+0x1c/0x29
[<c011776e>] try_to_wake_up+0x355/0x35f
[<c0116d0a>] find_busiest_group+0x177/0x46a
[<c01f9d21>] flush_to_ldisc+0x104/0x15c
[<f8f0a335>] mxser_receive_chars+0x241/0x249 [mxser_new]
[<f8f0ad7a>] mxser_interrupt+0x15f/0x1e6 [mxser_new]
[<c013fb83>] handle_IRQ_event+0x23/0x49
[<c013fc3c>] __do_IRQ+0x93/0xe8
[<c01050e5>] do_IRQ+0x43/0x52
[<c01036b6>] common_interrupt+0x1a/0x20
Code: 83 e0 ee 89 41 3c 25 ee 00 00 00 42 eb 19 0f b6 42 1a 8b 51 08 89 41 38 31 c0 42 ee 89 d8 83 c8 02 89 41 3c 0f b6 c0 ee 8b 41 04 <8b> 40 68 83 78 08 00 79 15 8b 41 40 8b 51 08 83 e0 fd 89 41 40
EIP: [<f8f0911f>] mxser_stoprx+0x54/0x74 [mxser_new] SS:ESP 0068:ec901e20
<0>Kernel panic - not syncing: Fatal exception in interrupt-- Sergei.
Quoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
I've tested the latest version you've sent me yesterday. Still no luck. The oops is below. I'll try with low_latency commented in a few minutes.
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000068 printing eip: f8f3711f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss i2c_i801 psmouse snd_pcm floppy tsdev serio_raw snd_timer parport_pc parport snd i2c_core pcspkr evdev soundcore snd_page_alloc mxser_new rtc ext3 jbd mbcache sd_mod ide_cd cdrom usb_storage usbhid ata_piix libata piix scsi_mod generic uhci_hcd ide_core usbcore skge thermal processor fan CPU: 0 EIP: 0060:[<f8f3711f>] Tainted: P VLI EFLAGS: 00010246 (2.6.18-3-686 #1)
EIP is at mxser_stoprx+0x54/0x74 [mxser_new]
eax: 00000000 ebx: 00000001 ecx: f8f3d79c edx: dfde7240
esi: f391211b edi: f4f83fff ebp: c1b17800 esp: f5bc7d7c
ds: 007b es: 007b ss: 0068
Process make (pid: 17088, ti=f5bc6000 task=f4febaa0 task.ti=f5bc6000)
Stack: 00000001 c01fe4af 000000ff f391201c c1b17c04 c011669e 00000000 00000000
00000003 00000096 c1b1780c 00000286 00000001 00000096 c01f97c9 00000000
f456f800 00000286 f8f38335 f5bc7e14 f8f3d8a8 00000286 c01f9806 f456f800
Call Trace:
[<c01fe4af>] n_tty_receive_buf+0xcd6/0xcf9
[<c011669e>] __wake_up+0x2a/0x3d
[<c01f97c9>] tty_ldisc_deref+0x50/0x5f
[<f8f38335>] mxser_receive_chars+0x241/0x249 [mxser_new]
[<c01f9806>] tty_ldisc_try+0x2e/0x32
[<c01f9d4c>] flush_to_ldisc+0x12f/0x15c
[<c01f9793>] tty_ldisc_deref+0x1a/0x5f
[<f8f38335>] mxser_receive_chars+0x241/0x249 [mxser_new]
[<f8f37678>] mxser_transmit_chars+0x14/0x164 [mxser_new]
[<c01f9d21>] flush_to_ldisc+0x104/0x15c
[<f8f38335>] mxser_receive_chars+0x241/0x249 [mxser_new]
[<f8f38d7a>] mxser_interrupt+0x15f/0x1e6 [mxser_new]
[<c013fb83>] handle_IRQ_event+0x23/0x49
[<c013fc3c>] __do_IRQ+0x93/0xe8
[<c01050e5>] do_IRQ+0x43/0x52
[<c01036b6>] common_interrupt+0x1a/0x20
[<c01bacf4>] __copy_from_user_ll+0x19/0x38
[<c01071de>] convert_fxsr_from_user+0x15/0xd5
[<c0164bc8>] pipe_read+0x0/0x1e
[<c010774e>] restore_i387+0x6f/0xc0
[<c0102203>] restore_sigcontext+0x10f/0x165
[<c0102b3d>] sys_sigreturn+0xa4/0xcb
[<c0102c7b>] syscall_call+0x7/0xb
Code: 83 e0 ee 89 41 3c 25 ee 00 00 00 42 eb 19 0f b6 42 1a 8b 51 08 89 41 38 31 c0 42 ee 89 d8 83 c8 02 89 41 3c 0f b6 c0 ee 8b 41 04 <8b> 40 68 83 78 08 00 79 15 8b 41 40 8b 51 08 83 e0 fd 89 41 40
EIP: [<f8f3711f>] mxser_stoprx+0x54/0x74 [mxser_new] SS:ESP 0068:f5bc7d7c
<0>Kernel panic - not syncing: Fatal exception in interruptQuoted 11 lines · Jiri Slaby
Quoted 8 lines · [email protected] wrote:
Hi Jiri,
I've figured out that both old and new mxser drivers have two similar problems:
1. When there are data coming to a port, sometimes opening of the port entirely locks the box. This is quite reproducible. Any idea what's wrong and how can I help to debug it?
Could you test the patch below, if something changes?
Just tested with low_latency commented out. Still oopses:
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000008 printing eip: f8f1730f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer i2c_i801 tsdev psmouse snd soundcore snd_page_alloc i2c_core serio_raw parport_pc parport mxser_new evdev floppy pcspkr rtc ext3 jbd mbcache usb_storage usbhid ide_cd cdrom sd_mod uhci_hcd piix usbcore skge ata_piix libata scsi_mod generic ide_core thermal processor fan CPU: 0 EIP: 0060:[<f8f1730f>] Tainted: P VLI EFLAGS: 00010046 (2.6.18-3-686 #1)
EIP is at mxser_receive_chars+0x21b/0x249 [mxser_new]
eax: 00000000 ebx: 00000000 ecx: 00000000 edx: 00000286
esi: f8f1c79c edi: 00000001 ebp: c1be6000 esp: c0313efc
ds: 007b es: 007b ss: 0068
Process swapper (pid: 0, ti=c0312000 task=c02c76a0 task.ti=c0312000)
Stack: c0313f48 f8f1c8a8 0000903d 00000000 00000fff 000000ff 00000286 0000007b
f8f1c79c 00000006 0000007f 000000c6 f8f17d7a 00000007 00000008 00000080
00000000 00000000 f8f1bdc0 00000060 df985500 00000000 00000000 0000003a
Call Trace:
[<f8f17d7a>] mxser_interrupt+0x15f/0x1e6 [mxser_new]
[<c013fb83>] handle_IRQ_event+0x23/0x49
[<c013fc3c>] __do_IRQ+0x93/0xe8
[<c01050e5>] do_IRQ+0x43/0x52
[<c01036b6>] common_interrupt+0x1a/0x20
[<c0101b91>] mwait_idle+0x25/0x38
[<c0101b52>] cpu_idle+0x9f/0xb9
[<c03186fd>] start_kernel+0x379/0x380
Code: ed ff ff eb 1f 8b 06 83 78 18 00 75 17 8b 56 08 83 c2 05 ec 8b 14 24 0f b6 c0 a8 01 89 02 0f 85 d4 fe ff ff 8b 46 04 8b 54 24 18 <8b> 40 08 01 3c 85 c4 ec f1 f8 8b 44 24 04 01 be f4 00 00 00 01
EIP: [<f8f1730f>] mxser_receive_chars+0x21b/0x249 [mxser_new] SS:ESP 0068:c0313efc
<0>Kernel panic - not syncing: Fatal exception in interrupt
BUG: warning at arch/i386/kernel/smp.c:547/smp_call_function()
[<c010f5a3>] smp_call_function+0x53/0xfe
[<c011d97e>] printk+0x14/0x18
[<c010f661>] smp_send_stop+0x13/0x1c
[<c011cfc6>] panic+0x4c/0xe2
[<c0104013>] die+0x256/0x28a
[<c01156e0>] do_page_fault+0x3b4/0x481
[<c01f9f8d>] tty_buffer_request_room+0x107/0x112
[<c011532c>] do_page_fault+0x0/0x481
[<c01037f9>] error_code+0x39/0x40
[<f8f1730f>] mxser_receive_chars+0x21b/0x249 [mxser_new]
[<f8f17d7a>] mxser_interrupt+0x15f/0x1e6 [mxser_new]
[<c013fb83>] handle_IRQ_event+0x23/0x49
[<c013fc3c>] __do_IRQ+0x93/0xe8
[<c01050e5>] do_IRQ+0x43/0x52
[<c01036b6>] common_interrupt+0x1a/0x20
[<c0101b91>] mwait_idle+0x25/0x38
[<c0101b52>] cpu_idle+0x9f/0xb9
[<c03186fd>] start_kernel+0x379/0x380Quoted 16 lines
Quoted 1 line · Jiri Slaby
Could you test the patch below, if something changes?
Just tested with
low_latencycommented out. Still oopses:BUG: unable to handle kernel NULL pointer dereference at virtual address 00000008 printing eip: f8f1730f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer i2c_i801 tsdev psmouse snd soundcore snd_page_alloc i2c_core serio_raw parport_pc parport mxser_new evdev floppy pcspkr rtc ext3 jbd mbcache usb_storage usbhid ide_cd cdrom sd_mod uhci_hcd piix usbcore skge ata_piix libata scsi_mod generic ide_core thermal processor fan CPU: 0 EIP: 0060:[<f8f1730f>] Tainted: P VLI EFLAGS: 00010046 (2.6.18-3-686 #1)
Quoted 1 line
EIP is at
mxser_receive_chars+0x21b/0x249[mxser_new]
Yes, port->tty still somewhere becomes NULL -- does this patch help?
---
mxser_new.c +6 -3
@@ -2051,13 +2051,16 @@ static void mxser_wait_until_sent(structvoid mxser_hangup (struct tty_struct *tty) {struct mxser_port *info = tty->driver_data; +unsigned long flags;mxser_flush_buffer (tty);mxser_shutdown (info); +spin_lock_irqsave (&info->slock, flags); info->event =0 ; info->count =0 ; info->flags &= ~ASYNC_NORMAL_ACTIVE ; info->tty =NULL ; +spin_unlock_irqrestore (&info->slock, flags);wake_up_interruptible (&info->open_wait); }
@@ -2263,8 +2266,6 @@ static irqreturn_t mxser_interrupt(int iif (bits & irqbits)continue ; port = &brd->ports[i]; -if (!(port->flags &ASYNC_INITIALIZED )) -continue ; int_cnt =0 ;spin_lock (&port->slock);
@@ -2274,7 +2275,9 @@ static irqreturn_t mxser_interrupt(int ibreak ; iir &=MOXA_MUST_IIR_MASK ;if (!port->tty || - (port->flags &ASYNC_CLOSING )) { + (port->flags &ASYNC_CLOSING ) || + !(port->flags & +ASYNC_INITIALIZED )) { status =inb (port->ioaddr +UART_LSR );outb (0x27 , port->ioaddr +UART_FCR );inb (port->ioaddr +UART_MSR );
Quoted 18 lines · Jiri Slaby
Quoted 16 lines
Quoted 1 line · Jiri Slaby
Could you test the patch below, if something changes?
Just tested with
low_latencycommented out. Still oopses:BUG: unable to handle kernel NULL pointer dereference at virtual address 00000008 printing eip: f8f1730f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer i2c_i801 tsdev psmouse snd soundcore snd_page_alloc i2c_core serio_raw parport_pc parport mxser_new evdev floppy pcspkr rtc ext3 jbd mbcache usb_storage usbhid ide_cd cdrom sd_mod uhci_hcd piix usbcore skge ata_piix libata scsi_mod generic ide_core thermal processor fan CPU: 0 EIP: 0060:[<f8f1730f>] Tainted: P VLI EFLAGS: 00010046 (2.6.18-3-686 #1)Quoted 1 line
EIP is at
mxser_receive_chars+0x21b/0x249[mxser_new]
Quoted 1 line
Yes,
port->ttystill somewhere becomes NULL -- does this patch help?
No, it still oopses (the oops is at the end).
In addition, before this latest patch was applied, when playing with two simultaneously run programs that do nothing but open/close the port in a loop (and the port was idle, i.e., nothing came to it from outside), I've once got this thing:
irq 58: nobody cared (try booting with the "irqpoll" option)
[<c014037f>] __report_bad_irq+0x2b/0x69 [<c014056c>] note_interrupt+0x1af/0x1e7 [<c013fb83>] handle_IRQ_event+0x23/0x49 [<c013fc5c>] __do_IRQ+0xb3/0xe8 [<c01050e5>] do_IRQ+0x43/0x52 [<c01036b6>] common_interrupt+0x1a/0x20 [<c011007b>] __cpu_up+0x88/0x164 [<c02810a0>] _spin_unlock_irqrestore+0x8/0x9 [<f8f8bf57>] mxser_startup+0x177/0x190 [mxser_new] [<f8f8c3da>] mxser_open+0x6b/0x251 [mxser_new] [<c01fcafd>] tty_open+0x16a/0x2e8 [<c01619a1>] chrdev_open+0x126/0x141 [<c016187b>] chrdev_open+0x0/0x141 [<c0158cb1>] __dentry_open+0xc8/0x1ac [<c0158df9>] nameidata_to_filp+0x19/0x28 [<c0158e33>] do_filp_open+0x2b/0x31 [<c0158e77>] do_sys_open+0x3e/0xb3 [<c0158f19>] sys_open+0x16/0x18 [<c0102c11>] sysenter_past_esp+0x56/0x79
handlers:
[<f8f8cc1b>] (mxser_interrupt+0x0/0x1e6 [mxser_new]) Disabling IRQ #58
Another thing that I've noticed, is that when two programs that open/close a port in a loop run simultaneously, each of them begin to sometimes fail to open the port with errno=5 (I/O Errror). Though this thing happends even if I try to use /dev/ttyS0, so it is not mxser-specific (/dev/null never fails).
Here is the oops I promised at the beginning:
BUG: unable to handle kernel NULL pointer dereference at virtual address 00000008 printing eip: f927630f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: mxser_new nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm psmouse snd_timer i2c_i801 snd serio_raw i2c_core tsdev parport_pc parport pcspkr evdev rtc soundcore snd_page_alloc floppy ext3 jbd mbcache sd_mod usb_storage usbhid ide_cd cdrom ata_piix libata scsi_mod uhci_hcd piix generic skge usbcore ide_core thermal processor fan CPU: 0 EIP: 0060:[<f927630f>] Tainted: P VLI EFLAGS: 00010046 (2.6.18-3-686 #1)
EIP is at mxser_receive_chars+0x21b/0x249 [mxser_new]
eax: 00000000 ebx: 00000000 ecx: 00000000 edx: 00000286
esi: f927b79c edi: 00000001 ebp: c1991800 esp: c0313efc
ds: 007b es: 007b ss: 0068
Process swapper (pid: 0, ti=c0312000 task=c02c76a0 task.ti=c0312000)
Stack: c0313f48 f927b8a8 0000903a 00000000 00000fff 000000ff 00000286 0000007b
f927b79c 00000006 0000007f 000000c6 f9276d7a 00000007 00000008 00000080
00000000 00000000 f927adc0 00000060 dfc74980 00000000 00000000 0000003a
Call Trace:
[<f9276d7a>] mxser_interrupt+0x15f/0x1e6 [mxser_new]
[<c013fb83>] handle_IRQ_event+0x23/0x49
[<c013fc3c>] __do_IRQ+0x93/0xe8
[<c01050e5>] do_IRQ+0x43/0x52
[<c01036b6>] common_interrupt+0x1a/0x20
[<c0101b91>] mwait_idle+0x25/0x38
[<c0101b52>] cpu_idle+0x9f/0xb9
[<c03186fd>] start_kernel+0x379/0x380
Code: ed ff ff eb 1f 8b 06 83 78 18 00 75 17 8b 56 08 83 c2 05 ec 8b 14 24 0f b6 c0 a8 01 89 02 0f 85 d4 fe ff ff 8b 46 04 8b 54 24 18 <8b> 40 08 01 3c 85 c4 dc 27 f9 8b 44 24 04 01 be f4 00 00 00 01
EIP: [<f927630f>] mxser_receive_chars+0x21b/0x249 [mxser_new] SS:ESP 0068:c0313efc
<0>Kernel panic - not syncing: Fatal exception in interrupt
BUG: warning at arch/i386/kernel/smp.c:547/smp_call_function()
[<c010f5a3>] smp_call_function+0x53/0xfe
[<c011d97e>] printk+0x14/0x18
[<c010f661>] smp_send_stop+0x13/0x1c
[<c011cfc6>] panic+0x4c/0xe2
[<c0104013>] die+0x256/0x28a
[<c01156e0>] do_page_fault+0x3b4/0x481
[<c01f9f8d>] tty_buffer_request_room+0x107/0x112
[<c011532c>] do_page_fault+0x0/0x481
[<c01037f9>] error_code+0x39/0x40
[<f927630f>] mxser_receive_chars+0x21b/0x249 [mxser_new]
[<f9276d7a>] mxser_interrupt+0x15f/0x1e6 [mxser_new]
[<c013fb83>] handle_IRQ_event+0x23/0x49
[<c013fc3c>] __do_IRQ+0x93/0xe8
[<c01050e5>] do_IRQ+0x43/0x52
[<c01036b6>] common_interrupt+0x1a/0x20
[<c0101b91>] mwait_idle+0x25/0x38
[<c0101b52>] cpu_idle+0x9f/0xb9
[<c03186fd>] start_kernel+0x379/0x380Quoted 18 lines · Jiri Slaby
Quoted 16 lines
Quoted 1 line · Jiri Slaby
Could you test the patch below, if something changes?
Just tested with
low_latencycommented out. Still oopses:BUG: unable to handle kernel NULL pointer dereference at virtual address 00000008 printing eip: f8f1730f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: nvidia agpgart ipv6 nfs lockd nfs_acl sunrpc dm_mod sr_mod sbp2 ieee1394 ide_generic ide_disk e1000 snd_hda_intel snd_hda_codec snd_pcm_oss snd_mixer_oss snd_pcm snd_timer i2c_i801 tsdev psmouse snd soundcore snd_page_alloc i2c_core serio_raw parport_pc parport mxser_new evdev floppy pcspkr rtc ext3 jbd mbcache usb_storage usbhid ide_cd cdrom sd_mod uhci_hcd piix usbcore skge ata_piix libata scsi_mod generic ide_core thermal processor fan CPU: 0 EIP: 0060:[<f8f1730f>] Tainted: P VLI EFLAGS: 00010046 (2.6.18-3-686 #1)Quoted 1 line
EIP is at
mxser_receive_chars+0x21b/0x249[mxser_new]
Quoted 1 line
Yes,
port->ttystill somewhere becomes NULL -- does this patch help?
In addition to my previous answer. I've performed the same tests with regular PC serial port. The issues with "Disabling Irq #N" and with I/O error on open() exist with this port as well, so they aren't mxser-specific. I wasn't able to reproduce oopses, so they do seem to be mxser-specific.
-- Sergei.
Quoted 18 lines · Jiri Slaby
Quoted 16 lines
Quoted 1 line · Jiri Slaby
Could you test the patch below, if something changes?
Just tested with
low_latencycommented out. Still oopses:BUG: unable to handle kernel NULL pointer dereference at virtual address 00000008 printing eip: f8f1730f *pde = 00000000 Oops: 0000 [#1] SMP Modules linked in: ... CPU: 0 EIP: 0060:[<f8f1730f>] Tainted: P VLI EFLAGS: 00010046 (2.6.18-3-686 #1)Quoted 1 line
EIP is at
mxser_receive_chars+0x21b/0x249[mxser_new]
Quoted 1 line
Yes,
port->ttystill somewhere becomes NULL -- does this patch help?
Hi, Jiri!
I'm so sorry, I don't know what I smoked yesterday, but the latest version you've sent me does not have this problem anymore! I think I copied compiled module to modules directory for different kernel and thus tested old code all the time. BTW, should you tell me that the ports are now called /dev/ttyMIx instead of /dev/ttyMx, I think I'd notice my mistake earlier. Yes, in fact I didn't test any version where ports are called /dev/ttyMIx until now! In particular, it means that maybe some of the recent changes you've made yesterday based on my wrong reports are not in fact required.
Anyway, the only mxser-specific problem that currently remains for me and that I didn't see before, is the following:
# rmmod mxser_new Trying to free already-free IRQ 58 Trying to free nonexistent resource <0000000000009000-000000000000903f> Trying to free nonexistent resource <0000000000008800-0000000000008800> #
-- Sergei.
Quoted 1 line · Sergei Organov
Hi, Jiri!
Hi.
Quoted 2 lines
I'm so sorry, I don't know what I smoked yesterday, but the latest version you've sent me does not have this problem anymore! I think I
YES!
Quoted 3 lines
copied compiled module to modules directory for different kernel and thus tested old code all the time. BTW, should you tell me that the ports are now called
/dev/ttyMIxinstead of/dev/ttyMx, I think I'd
ttyM was reserved for isicom, and it caused many warnings in the kernel, when both isicom and mxser were built and loaded. The proper name for mxser is (and ever was) ttyMI -- sorry for not giving you a notice (I didn't realize the change).
Quoted 4 lines
notice my mistake earlier. Yes, in fact I didn't test any version where ports are called
/dev/ttyMIxuntil now! In particular, it means that maybe some of the recent changes you've made yesterday based on my wrong reports are not in fact required.
I think those with ASYNC_CLOSING test in the isr is the one (but also wakeup spinlock change is requisite to go upstream).
Quoted 7 lines
Anyway, the only mxser-specific problem that currently remains for me and that I didn't see before, is the following:
# rmmod mxser_new Trying to free already-free IRQ 58 Trying to free nonexistent resource <0000000000009000-000000000000903f> Trying to free nonexistent resource <0000000000008800-0000000000008800>
Thanks, I'll fix this and let you know. Does this happed every time you try to unload it?
thanks,
Signature · 4 lines
http://www.fi.muni.cz/~xslaby/ Jiri Slaby faculty of informatics, masaryk university, brno, cz e-mail: jirislaby gmail com, gpg pubkey fingerprint: B674 9967 0407 CE62 ACC8 22A0 32CC 55C3 39D4 7A7E
Jiri Slaby <[email protected]> writes:
[...]
Quoted 7 lines
Quoted 4 lines
# rmmod mxser_new Trying to free already-free IRQ 58 Trying to free nonexistent resource <0000000000009000-000000000000903f> Trying to free nonexistent resource <0000000000008800-0000000000008800>Thanks, I'll fix this and let you know. Does this happed every time you try to unload it?
Yes, it's stable. Happens every time.
-- Sergei.