malloc problem in file io ?

[Cpasjuste]

Hi,

When developing my usbhostfs module i noticed some random crash on file io. I thought the problem was in my module, but i isolated the problem to the malloc calls in "file.c". I guess you may pushing the limits there. You may be able to reproduce the crash by doing multiple times this procedure :

• copy directory "ux0:user" to "ux0:temp" (as an example)
• delete the directory "ux0:temp"

Before posting this report i tried disabling the only user (sceshell) module i use (psp2shell), whith the same results.

[Cpasjuste]

Doing more research i found that around 106 MB of memory is available, so i guess there's something more. I wonder if using malloc in threads (copy_thread, delete_thread..) is safe ?

[Cpasjuste]

Hi,

I'm now not sure it's a bug in VitaShell anymore :/
I was not able to reproduce the problem when not enabling my (kernel) usbhostfs plugin, so it may corrupt the memory (?!). I was almost sure to have this problem with no plugin enabled, but maybe i did something wrong. I'll try to take another look today and report. (by the way, malloc IS thread safe).

[Cpasjuste]

So, i can confirm the problem still happen randomly even with no plugins enabled (on enso at least). I was sometime able to do the copy more than 5 times without VitaShell crashing, sometime it crash at the first copy. It can also happen on delete.

[TheOfficialFloW]

please test 1.74

[limyz]

This problem is so similar to the one most recently posted - shall wait before tagging that as duplicate.

[Cpasjuste]

I will take a look tomorow.

[Cpasjuste]

From my testing i don't spot the problem with version 1.74. I still have random crashes when doing file io on host0 with my usbhostfs plugin, but i'm now sure it's a problem with my module. I guess you suspect the changes in the dialog code to cause this in 1.75. I would need to do more testing on 1.75 to be sure i didn't do something wrong when testing on it.

[Cpasjuste]

So, it seems i fixed my module, and there is no more problem with 1.74. I'll test again with 1.75 soon.

[Cpasjuste]

I can confirm the problem doesn't exist on 1.74 but exist on 1.75. I guess there is maybe a problem with this commit : e3f52cc

[Cpasjuste]

I guess there is a logic problem at this line, i'll take a closer look tomorrow :

VitaShell/uncommon_dialog.c

Line 76 in e3f52cc

tempstr[i+1] = '\0';

[Cpasjuste]

Also, strdup should be freed, which is not the case here.

VitaShell/uncommon_dialog.c

Line 75 in e3f52cc

tempstr = strdup(uncommon_dialog.msg);

[Cpasjuste]

Fixed in dae6212

[TheOfficialFloW]

Did you test the new build?

[Cpasjuste]

Yes just now, no more crash. I saw your changes in "dialog_uncommon", the problem was there for sure.

[limyz]

Awesome work