Workaround for broken Eaton (and others?) string language ID descriptor by tormodvolden · Pull Request #2604 · networkupstools/nut

tormodvolden · 2024-08-26T20:00:13Z

@masterwishx has tested this with his Eaton device, both on libusb0 and libusb1. Thanks!

The first two commits revert something that was an unsuccessful attempt at fixing something that turned out to be a buggy device. A more effective workaround is suggested in commits 4-5, which also tidies up and unifies things. Commit 3 reverts "retrials" for easier refactoring in 4-5 and similar functionality is added back in a more unified way in commit 4. Commits 6-7-8 are just tidying up things for future work, not strictly needed.

Fixes #1925
References #414

TODOs fixed in rework:

retries should be in nut_get_string and not in API wrapper
usb1 *_simple should just be kept as simple libusb0/1 API wrapper (if needed at all)

Ideas for future work in other PRs:

all drivers should use "nut_get_string" instead of *_simple and *_ascii
langid should be cached somewhere (in USBDevice_t or a new USBHandle_t ?)

masterwishx · 2024-08-26T23:21:56Z

@masterwishx Can you please test this with your Eaton device?

Wow , Thanks a lot , sure i will test it ...

masterwishx · 2024-08-27T07:39:47Z

I will try to compile without unraid nut plugin as author of plugin don't have time for it...

masterwishx · 2024-08-27T08:22:26Z

Can I test now?

tormodvolden · 2024-08-27T08:24:27Z

Yes. I had only done minor/optical refinements.

masterwishx · 2024-08-27T12:25:12Z

@tormodvolden Thanks a lot , YES, after compiled i have all info :

@jimklimov just battery.voltage and battery.voltage.nominal is missing now :(

battery.capacity: 7.00
battery.charge: 100
battery.charge.low: 20
battery.charge.restart: 0
battery.protection: yes
battery.runtime: 3399
battery.runtime.low: 180
battery.type: PbA                         <  ----   wasnt avalible befor !!!
device.mfr: EATO                           <  ----   wasnt avalible befor !!!  EATON should be ? 
device.model: Eaton 9 2000      <  ----   wasnt avalible befor !!!
device.serial:   xxxxxxxxxxx        #REDACTED by me # !!! serial is shown now !!! 
device.type: ups
driver.debug: 0
driver.flag.allow_killpower: 0
driver.name: usbhid-ups
driver.parameter.pollfreq: 30
driver.parameter.pollinterval: 2
driver.parameter.port: auto
driver.parameter.synchronous: auto
driver.state: updateinfo
driver.version: 2.8.2.1
driver.version.data: MGE HID 1.48
driver.version.internal: 0.56
driver.version.usb: libusb-1.0.26 (API: 0x1000108)
input.bypass.frequency: 50.0
input.bypass.voltage: 232.0
input.frequency: 50.0
input.frequency.nominal: 50
input.transfer.high: 300
input.transfer.low: 100
input.voltage: 232.0
input.voltage.nominal: 233
outlet.1.status: on
outlet.desc: Main Outlet
outlet.id: 0
outlet.switchable: no
output.current: 1.60
output.frequency: 50.0
output.frequency.nominal: 50
output.voltage: 230.0
output.voltage.nominal: 230
ups.beeper.status: enabled
ups.delay.shutdown: 20
ups.delay.start: 30
ups.firmware: 00.02.00    <  ----   wasnt avalible befor !!!
ups.load: 20
ups.load.high: 105
ups.mfr: EATO                   <  ----   wasnt avalible befor !!!    EATON should be ?
ups.model: Eaton 9 2000         <  ----   wasnt avalible befor !!!    
ups.power: 398
ups.power.nominal: 2000
ups.productid: ffff
ups.realpower: 265
ups.realpower.nominal: 1600
ups.serial: xxxxxxxxxxxx                   #REDACTED by me # !!! serial is shown now !!! 
ups.shutdown: enabled
ups.start.auto: yes
ups.start.battery: yes
ups.start.reboot: yes
ups.status: OL
ups.temperature: 24.9
ups.test.interval: 604800
ups.test.result: Done and passed
ups.timer.shutdown: -1
ups.timer.start: -1
ups.type: online
ups.vendorid: 0463

masterwishx · 2024-08-27T12:35:54Z

Also maybe Eaton 9E 2000i should be instead of Eaton 9 2000 ... and PbAc instead of PbA ...

it seems like the last character is missing in values, also in serial number

tormodvolden · 2024-08-27T12:42:54Z

Thanks for your testing. I probably made an off-by-one error in the string decoding. I'll fix that later.

tormodvolden · 2024-08-27T12:49:12Z

The missing last character should be fixed now.

masterwishx · 2024-08-27T13:16:30Z

The missing last character should be fixed now.

Now its fine and all values avaliable :

battery.capacity: 7.00
battery.charge: 100
battery.charge.low: 20
battery.charge.restart: 0
battery.protection: yes
battery.runtime: 3444
battery.runtime.low: 180
battery.type: PbAc
battery.voltage: 83.0
battery.voltage.nominal: 72
device.mfr: EATON
device.model: 9E2000i
device.serial:  xxxxxxxxxxxx                   #REDACTED by me # !!! serial is shown now !!!
device.type: ups
driver.debug: 0
driver.flag.allow_killpower: 0
driver.name: usbhid-ups
driver.parameter.pollfreq: 30
driver.parameter.pollinterval: 2
driver.parameter.port: auto
driver.parameter.synchronous: auto
driver.state: updateinfo
driver.version: 2.8.2.1
driver.version.data: MGE HID 1.48
driver.version.internal: 0.56
driver.version.usb: libusb-1.0.26 (API: 0x1000108)
input.bypass.frequency: 49.9
input.bypass.voltage: 232.0
input.frequency: 49.9
input.frequency.nominal: 50
input.transfer.high: 300
input.transfer.low: 100
input.voltage: 232.0
input.voltage.nominal: 231
outlet.1.status: on
outlet.desc: Main Outlet
outlet.id: 0
outlet.switchable: no
output.current: 1.60
output.frequency: 49.9
output.frequency.nominal: 50
output.voltage: 229.0
output.voltage.nominal: 230
ups.beeper.status: enabled
ups.delay.shutdown: 20
ups.delay.start: 30
ups.firmware: 00.02.000
ups.load: 19
ups.load.high: 105
ups.mfr: EATON
ups.model: 9E2000i
ups.power: 393
ups.power.nominal: 2000
ups.productid: ffff
ups.realpower: 261
ups.realpower.nominal: 1600
ups.serial:  xxxxxxxxxxxx                   #REDACTED by me # !!! serial is shown now !!!
ups.shutdown: enabled
ups.start.auto: yes
ups.start.battery: yes
ups.start.reboot: yes
ups.status: OL
ups.temperature: 24.9
ups.test.interval: 604800
ups.test.result: Done and passed
ups.timer.shutdown: -1
ups.timer.start: -1
ups.type: online
ups.vendorid: 0463

masterwishx · 2024-08-27T15:03:57Z

@jimklimov when you will have time, please check it, the issue is solved :)

jimklimov · 2024-08-27T15:33:30Z

Sounds great, thanks a lot! I'll take a closer look later though.

Interesting how battery.voltage(.nominal) was missing and came back after the off-by-one error fix. Was it supposed to be impacted? Not a string, right?..

jimklimov · 2024-08-27T15:34:16Z

@tormodvolden : do you expect this fix is generic enough to drop the need for older *langid* options?

masterwishx · 2024-08-28T07:58:43Z

Sounds great, thanks a lot! I'll take a closer look later though.

Interesting how battery.voltage(.nominal) was missing and came back after the off-by-one error fix. Was it supposed to be impacted? Not a string, right?..

I'm not sure but maybe battery.voltage(.nominal) related to your fix for Eaton 9E names, so when name was not fully then it was not catched?

masterwishx · 2024-08-28T08:01:14Z

@tormodvolden do you need to finish also TODO befor merging to main?

tormodvolden · 2024-08-28T16:19:54Z

@masterwishx I have reworked it to address the most important TODO items. It should now work equally well with libsb-1.0 and libusb0 so I would appreciate if you can test both.

@jimklimov The USB code in nut seems complicated by the burden of supporting both libusb0 and libusb1. What are the reasons for supporting libusb0?

There seems to be drivers using the libusb0.c and libusb1.c through the "API" and there seems to be some drivers that do their own stuff. And an outlier tools/nut-scanner/scan_usb.c which does its own dlopen() fun and supports both. Then there are wrappers trying to unify the APIs but not being consistently used. Unifying all this and tidying up would probably help in the long run but could be a lot of work.

I would suggest as a next step to get rid of all direct use of usb_get_string_simple() and libusb_get_string_descriptor_ascii() and instead use the "API" functions or the new common nut_usb_get_string().

tormodvolden · 2024-08-28T16:39:41Z

@tormodvolden : do you expect this fix is generic enough to drop the need for older *langid* options?

The fix in this PR doesn't honour the "langid_fix" option, but will always try to auto-detect by reading out the language ID. Having a way to override the fallback value of en_US could possibly be useful. This option should be a generic one, not limited to a single driver. I don't know if anyone used this option to set anything else than en_US (0x0409). If it only was used for setting 0x0409 and the affected devices (only seen in nutdrv_qx and blazer_usb) won't get hosed up by a langid request (but only return an invalid descriptor) then this option can be dropped now. EDIT: One device in data/driver.list.in uses 0x4095.

I am not sure the "noscanlangid" option adds much value, unless some broken device get hosed up by such requests. Other code is already "scanning" it (reading out the language ID), like is being done here in nut_usb_get_string() as well. The option might be just leftovers from previous work. Please check with those who use it (only seen in nutdrv_qx/hunnox). EDIT: The "hunnox" protocol seems to use a port-knocking scheme involving a sequence of non-standard string descriptors retrievals to unlock the device communication, so it has to do precise low-level string descriptor fetching, incompatible with the USB standard. Maybe nut_usb_get_string() will work fine after the unlocking.

masterwishx · 2024-08-28T17:09:41Z

I have reworked it to address the most important TODO items. It should now work equally well with libsb-1.0 and libusb0 so I would appreciate if you can test both.

Sure I can test it, but don't know how to test both?

jimklimov · 2024-08-28T17:21:51Z

Regarding libusb-0.1, on one hand this code was with NUT much longer (1.0 only getting into mainline a couple of years back with 2.8.0, so much less exposed and tested). On the other, NUT aims to please older platforms (and machines) where it once ran, and those may lack a build/package of libusb-1.0, I suppose.

tormodvolden · 2024-08-28T17:26:38Z

Sure I can test it, but don't know how to test both?

After a fresh checkout or running "make distclean":

mkdir build0 && cd build0 && ../configure --with-usb=libusb-0.1 && make
mkdir build1 && cd build1 && ../configure --with-usb=libusb-1.0 && make

jimklimov · 2024-08-28T17:27:18Z

As for testing, if the libusb-0.1-dev or similar is available, you can configure --with-libusb=0.1 && make

tormodvolden · 2024-08-28T17:29:23Z

Let me know if there is any platform where nut builds but there is no libusb-1.0.

masterwishx · 2024-08-28T18:04:34Z

Sure I can test it, but don't know how to test both?

After a fresh checkout or running "make distclean":
mkdir build0 && cd build0 && ../configure --with-usb=libusb-0.1 && make
mkdir build1 && cd build1 && ../configure --with-usb=libusb-1.0 && make

What I compiled in Slackware VM last time for test was nut.SlackBuild from Unraid NUT plugin modified by me. So it was libusb 1.0.26 that my Unraid version has.

So I can test latest PR tomorrow same way, but for libusb 0.1 I can add libusb-0.1 to .. /configure but will it work?

jimklimov · 2024-08-28T19:07:53Z

Let me know if there is any platform where nut builds but there is no libusb-1.0.

The oldest occasionally tested that I know of, because there still is practical interest, is Solaris 8. Gotta check if that involves USB, and which, though.

masterwishx · 2024-08-29T09:30:01Z

I have reworked it to address the most important TODO items. It should now work equally well with libsb-1.0 and libusb0 so I would appreciate if you can test both.

Compiled with libusb 1.0 , all working fine as before , all values available .

will check with libusb 0.1 as Unraid have libusb-compat 0.1.7

Fixes networkupstools#1925 Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

The string index is a number between 1 and 255, simply because it is the lower byte of wValue in the GET_DESCRIPTOR request. String index 0 is used for retrieving the langid array, so it is not valid for this function. There is no reason to impose a limit to the length of the buffer that the user is lending us, certainly not above the range of the involved types, so just a minimal sanity check is enough. Since in the minimal case we would return an empty zero-terminated string, it must at least have room for this. Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

The original names could also be confused with libusb-1.0 API functions. Also rename nut_libusb_strerror() although not a nut API function. Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

…s#1925, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

masterwishx · 2024-09-03T16:53:11Z

make[1]: *** [Makefile:1644: install-win-bundle-thirdparty] Error 1

make: *** [Makefile:1638: install-win-bundle] Error 2
Not sure what Appveyor disliked here, build/test were ok. So this fault is unrelated to this PR.

FWIW, it listed some DLLs that were not in mingw (FOSS) locations but that's normal (e.g. OS own libs). Not sure if the list is same as before, or is why the step failed. Maybe libexec/sockdebug.exe troubleshooting helper being not covered by DLL accompaniment?..

Also I saw it was fine in test befor...

AppVeyorBot · 2024-09-03T17:41:15Z

✅ Build nut 2.8.2.2109-master completed (commit 0511491169 by @jimklimov)

…t_libusb_set_altinterface() to maintain a consistent code style [networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

…rkupstools#1925, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

…)` method [networkupstools#1925, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

masterwishx · 2024-09-05T13:01:29Z

Finally merge is possible

Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

…on.c [networkupstools#2604, networkupstools#2615] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

…c::nut_usb_get_string(): trace the bytes going around [networkupstools#3201, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

tormodvolden force-pushed the eaton_broken_strings branch from 4eefec5 to a2475e8 Compare August 26, 2024 20:02

tormodvolden mentioned this pull request Aug 26, 2024

Eaton 9E2000I unknown device and ups model #1925

Closed

tormodvolden force-pushed the eaton_broken_strings branch from a2475e8 to 6d664a1 Compare August 26, 2024 22:00

tormodvolden force-pushed the eaton_broken_strings branch 2 times, most recently from 7d61ea3 to a3d2773 Compare August 27, 2024 08:16

tormodvolden force-pushed the eaton_broken_strings branch from 5b904b1 to 91fd69d Compare August 28, 2024 16:07

tormodvolden and others added 5 commits September 3, 2024 18:25

libusb0/1: Use nut_usb_get_string() instead of libusb functions

251202f

Fixes networkupstools#1925 Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

libusb0/1: Fix up comments about expected types

dc8bc69

Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

libusb0: Rename nut API functions for consistency with libusb1.c

42d66bd

The original names could also be confused with libusb-1.0 API functions. Also rename nut_libusb_strerror() although not a nut API function. Signed-off-by: Tormod Volden <debian.tormod@gmail.com>

NEWS.adoc: document new nut_usb_get_string() method [networkupstool…

f2ccec9

…s#1925, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

jimklimov added 3 commits September 3, 2024 22:14

drivers/libusb{0,1}.c: rename static nut_usb_set_altinterface() => nu…

a5df69f

…t_libusb_set_altinterface() to maintain a consistent code style [networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

NEWS.adoc: document new nut_usb_get_string() method - extend [netwo…

f202875

…rkupstools#1925, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

Bump USB-capable driver versions impacted by new `nut_usb_get_string(…

45c07f3

…)` method [networkupstools#1925, networkupstools#2604] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

tormodvolden force-pushed the eaton_broken_strings branch from d2b262c to 45c07f3 Compare September 3, 2024 20:24

jimklimov merged commit 6443582 into networkupstools:master Sep 5, 2024

tormodvolden deleted the eaton_broken_strings branch September 5, 2024 15:03

masterwishx mentioned this pull request Sep 6, 2024

Eaton 9E Model (bypass and Eco) Low/High Voltage Transfer is missing however present when usbhid-usb debug Driver #2492

Closed

jimklimov added a commit to jimklimov/nut that referenced this pull request Sep 11, 2024

drivers/usb-common.c: fix C style of comments [networkupstools#2604]

9f40a8f

Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

jimklimov added a commit to jimklimov/nut that referenced this pull request Sep 11, 2024

tools/nut-scanner/scan_usb.c: port nut_usb_get_string() from usb-comm…

938e80e

…on.c [networkupstools#2604, networkupstools#2615] Signed-off-by: Jim Klimov <jimklimov+nut@gmail.com>

jimklimov mentioned this pull request Sep 11, 2024

Port and use a variant of nut_usb_get_string() for nut-scanner #2626

Merged

jimklimov mentioned this pull request Sep 19, 2024

CI: Jenkins wants perhaps too many permissions from users #2634

Closed

masterwishx mentioned this pull request Nov 16, 2024

nut_libusb_get_interrupt: Connection timed out on debug #2644

Open

This was referenced Dec 16, 2024

usbhid-ups generally: apply disable_fix_report_desc…; cps-hid: fix mismatched LogMax between input/output voltages (bad encoding) #2718

Merged

CPS HID: Many unknown readings #2729

Open

jimklimov mentioned this pull request Jan 29, 2025

PR1500LCDRT2U Cyber Power UPS - serial number not reported so cannot run more than 1 UPS per server - tried device/port #2791

Closed

jimklimov mentioned this pull request Mar 20, 2025

Persistent usbhid-ups[7209]: nut_libusb_get_report: Input/Output Error #2854

Open

masterwishx mentioned this pull request Dec 7, 2025

Eaton 9E: ups.firmware variable reports a truncated firmware string #3201

Closed

Uh oh!

Conversation

tormodvolden commented Aug 26, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

masterwishx commented Aug 26, 2024

Uh oh!

masterwishx commented Aug 27, 2024

Uh oh!

masterwishx commented Aug 27, 2024

Uh oh!

tormodvolden commented Aug 27, 2024

Uh oh!

masterwishx commented Aug 27, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

masterwishx commented Aug 27, 2024

Uh oh!

tormodvolden commented Aug 27, 2024

Uh oh!

tormodvolden commented Aug 27, 2024

Uh oh!

masterwishx commented Aug 27, 2024

Uh oh!

masterwishx commented Aug 27, 2024

Uh oh!

jimklimov commented Aug 27, 2024

Uh oh!

jimklimov commented Aug 27, 2024

Uh oh!

masterwishx commented Aug 28, 2024

Uh oh!

masterwishx commented Aug 28, 2024

Uh oh!

tormodvolden commented Aug 28, 2024

Uh oh!

tormodvolden commented Aug 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

masterwishx commented Aug 28, 2024

Uh oh!

jimklimov commented Aug 28, 2024

Uh oh!

tormodvolden commented Aug 28, 2024

Uh oh!

jimklimov commented Aug 28, 2024

Uh oh!

tormodvolden commented Aug 28, 2024

Uh oh!

masterwishx commented Aug 28, 2024

Uh oh!

jimklimov commented Aug 28, 2024

Uh oh!

masterwishx commented Aug 29, 2024

Uh oh!

masterwishx commented Sep 3, 2024

Uh oh!

AppVeyorBot commented Sep 3, 2024

Uh oh!

masterwishx commented Sep 5, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

tormodvolden commented Aug 26, 2024 •

edited

Loading

masterwishx commented Aug 27, 2024 •

edited

Loading

tormodvolden commented Aug 28, 2024 •

edited

Loading