some remaining section mismatches
1. I don't understand this one. Can you look at it?
WARNING: drivers/video/macmodes.o - Section mismatch: reference to .init.text:mac_find_mode from __ksymtab between '__ksymtab_mac_find_mode' (at offset 0x0) and '__ksymtab_mac_map_monitor_sense'
2. This is requires either a whitelist addition or a change to the struct name in the driver:
WARNING: drivers/scsi/megaraid/megaraid_mbox.o - Section mismatch: reference to .init.text:megaraid_probe_one from .data between 'megaraid_pci_driver_g' (at offset 0x2e0) and 'megaraid_mbox_version'
3. drivers/char/tpm_infineon.c:
WARNING: drivers/char/tpm/tpm_infineon.o - Section mismatch: reference to .init.text:tpm_inf_pnp_probe from .data between 'tpm_inf_pnp' (at offset 0x18) and 'tpm_inf'
WARNING: drivers/char/tpm/tpm_infineon.o - Section mismatch: reference to .exit.text:tpm_inf_pnp_remove from .data between 'tpm_inf_pnp' (at offset 0x20) and 'tpm_inf'
I don't see a problem here, the driver just isn't using the whitelisted struct name(s).
4. drivers/rtc/: rtc-sysfs.c and rtc-test.c:
WARNING: drivers/rtc/rtc-sysfs.o - Section mismatch: reference to .init.text:rtc_sysfs_add_device from .data between 'rtc_sysfs_interface' (at offset 0x18) and 'rtc_attr_group'
This could be OK, hopefully someone can say yes/no.
WARNING: drivers/rtc/rtc-test.o - Section mismatch: reference to .exit.text:test_remove from .data between 'test_drv' (at offset 0x8) and 'dev_attr_irq'
Looks OK, struct name is just not in the whitelist. Changing the struct name removes the warning.
---
~Randy
Quoted 4 lines · Randy.Dunlap
1. I don't understand this one. Can you look at it?
WARNING:
drivers/video/macmodes.o- Section mismatch: reference to .init.text:mac_find_mode from__ksymtabbetween '__ksymtab_mac_find_mode' (at offset0x0) and '__ksymtab_mac_map_monitor_sense'
mac_find_mode is EXPORT_SYMBOL() and marked __init
This does not look correct to me. How do we know that it is valid to use this exported symbol?
Fix: do not mark mac_find_mode() as __init?
Quoted 4 lines
2. This is requires either a whitelist addition or a change to the struct name in the driver:
WARNING:
drivers/scsi/megaraid/megaraid_mbox.o- Section mismatch: reference to .init.text:megaraid_probe_one from .data between 'megaraid_pci_driver_g' (at offset0x2e0) and 'megaraid_mbox_version'
As indicated a rename would do the trick. s/megaraid_pci_driver_g/megaraid_pci_driver/g
The _g syntax does not look common and for this case we will not soften up the pattern to look for "driver" anywhere in the variable.
Quoted 7 lines
3.
drivers/char/tpm_infineon.c:WARNING:
drivers/char/tpm/tpm_infineon.o- Section mismatch: reference to .init.text:tpm_inf_pnp_probe from .data between 'tpm_inf_pnp' (at offset0x18) and 'tpm_inf'
WARNING:drivers/char/tpm/tpm_infineon.o- Section mismatch: reference to .exit.text:tpm_inf_pnp_remove from .data between 'tpm_inf_pnp' (at offset0x20) and 'tpm_inf'I don't see a problem here, the driver just isn't using the whitelisted struct name(s).
Agree.
s/\<tpm_inf_pnp\>/tpm_inf_driver/g
Quoted 5 lines
4.
drivers/rtc/:rtc-sysfs.candrtc-test.c:WARNING:
drivers/rtc/rtc-sysfs.o- Section mismatch: reference to .init.text:rtc_sysfs_add_device from .data between 'rtc_sysfs_interface' (at offset0x18) and 'rtc_attr_group'This could be OK, hopefully someone can say yes/no.
Look sane. The naming *_interface looks generic - so maybe we should whitelist it too?
Quoted 4 lines
WARNING:
drivers/rtc/rtc-test.o- Section mismatch: reference to .exit.text:test_remove from .data between 'test_drv' (at offset0x8) and 'dev_attr_irq'Looks OK, struct name is just not in the whitelist. Changing the struct name removes the warning.
grepping shows up that naming *_drv is quite common. So the better fix would probarly be to whitelist it too?
Sam
Quoted 10 lines · On Sun, 16 Apr 2006 01:36:27 +0200 Sam Ravnborg wrote:
Quoted 4 lines · Randy.Dunlap
1. I don't understand this one. Can you look at it?
WARNING:
drivers/video/macmodes.o- Section mismatch: reference to .init.text:mac_find_mode from__ksymtabbetween '__ksymtab_mac_find_mode' (at offset0x0) and '__ksymtab_mac_map_monitor_sense'
mac_find_modeisEXPORT_SYMBOL()and marked__init
This does not look correct to me. How do we know that it is valid to use this exported symbol?
Fix: do not markmac_find_mode()as__init?
IIRC, I saw 2 instances of exported __init code, which seems odd and dangerous to me, but I don't know of any rules against it...
Quoted 20 lines
Quoted 4 lines
2. This is requires either a whitelist addition or a change to the struct name in the driver:
WARNING:
drivers/scsi/megaraid/megaraid_mbox.o- Section mismatch: reference to .init.text:megaraid_probe_one from .data between 'megaraid_pci_driver_g' (at offset0x2e0) and 'megaraid_mbox_version'As indicated a rename would do the trick. s/megaraid_pci_driver_g/megaraid_pci_driver/g
The _g syntax does not look common and for this case we will not soften up the pattern to look for "driver" anywhere in the variable.Quoted 7 lines
3.
drivers/char/tpm_infineon.c:WARNING:
drivers/char/tpm/tpm_infineon.o- Section mismatch: reference to .init.text:tpm_inf_pnp_probe from .data between 'tpm_inf_pnp' (at offset0x18) and 'tpm_inf'
WARNING:drivers/char/tpm/tpm_infineon.o- Section mismatch: reference to .exit.text:tpm_inf_pnp_remove from .data between 'tpm_inf_pnp' (at offset0x20) and 'tpm_inf'I don't see a problem here, the driver just isn't using the whitelisted struct name(s).
Agree.
s/\<tpm_inf_pnp\>/tpm_inf_driver/g
Quoted 7 lines
Quoted 5 lines
4.
drivers/rtc/:rtc-sysfs.candrtc-test.c:WARNING:
drivers/rtc/rtc-sysfs.o- Section mismatch: reference to .init.text:rtc_sysfs_add_device from .data between 'rtc_sysfs_interface' (at offset0x18) and 'rtc_attr_group'This could be OK, hopefully someone can say yes/no.
Look sane. The naming
*_interfacelooks generic - so maybe we should whitelist it too?
I think so.
Quoted 6 lines
Quoted 4 lines
WARNING:
drivers/rtc/rtc-test.o- Section mismatch: reference to .exit.text:test_remove from .data between 'test_drv' (at offset0x8) and 'dev_attr_irq'Looks OK, struct name is just not in the whitelist. Changing the struct name removes the warning.
grepping shows up that naming
*_drvis quite common. So the better fix would probarly be to whitelist it too?
Likely. and I see probe_one in the symbol whitelist but not init_one, and we have lots of probe functions that end with "init_one". Maybe something to remember in case we need it.
---
~Randy