Skip to contentSkip to conversation

misc: pci_endpoint_test: Fix pci_endpoint_test_{copy,write,read}() panic

PATCH In review Shunsuke Mie 5 participants 10 messages Aug 16, 2022 – Aug 22, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 misc: pci_endpoint_test: Fix pci_endpoint_test_{copy,write,read}() panic

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +3 -3
index 8f786a225dcf..d45426a73396 100644
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;
 
 	use_dma = !!(param.flags & PCITEST_FLAGS_USE_DMA);
@@ -498,7 +498,7 @@ static bool pci_endpoint_test_write(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;
 
 	use_dma = !!(param.flags & PCITEST_FLAGS_USE_DMA);
@@ -596,7 +596,7 @@ static bool pci_endpoint_test_read(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;
 
 	use_dma = !!(param.flags & PCITEST_FLAGS_USE_DMA);
Signature · 1 line
2.17.1
#1 lore raw
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Or have a "validate all parameters" function that you do it in one place?

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

thanks,

greg k-h

#2 lore raw
Quoted 66 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Quoted 5 lines

Or have a "validate all parameters" function that you do it in one place?

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

This is bad, thanks Greg for spotting it.

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

Lorenzo

Quoted 4 lines

thanks,

greg k-h

#3 lore raw

Hi Lorenzo,

Quoted 70 lines · Lorenzo Pieralisi
Quoted 69 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Or have a "validate all parameters" function that you do it in one place?

Quoted 7 lines
Quoted 2 lines

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c
This is bad, thanks Greg for spotting it.

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

The userspace tool (pcitest [1]) that calls these IOCTLS prints "NOT OKAY" on 0 and "OKAY" for 1.

I'll change to negative value for error and still keep '0' for data mismatch error?

[1] ->
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c

Thanks,
Kishon

Quoted 7 lines

Lorenzo

Quoted 4 lines

thanks,

greg k-h

#4 lore raw
Quoted 78 lines · Kishon Vijay Abraham I

Hi Lorenzo,

Quoted 75 lines · Lorenzo Pieralisi
Quoted 66 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Quoted 5 lines

Or have a "validate all parameters" function that you do it in one place?

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c
This is bad, thanks Greg for spotting it.

Quoted 8 lines
Quoted 2 lines

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

The userspace tool (pcitest [1]) that calls these IOCTLS prints "NOT OKAY" on 0 and "OKAY" for 1.

I'll change to negative value for error and still keep '0' for data mismatch error?

No, 0 as a return value from a system call means "success". You can not redefine it to mean something else for a tiny single kernel driver, otherwise so much for consistency.

0 is success, -ERROR_NUMBER is an error, please read the ioctl man page.

thanks,

greg k-h

#5 lore raw

+Mani

Quoted 85 lines · Greg Kroah-Hartman
Quoted 84 lines · Kishon Vijay Abraham I

Hi Lorenzo,

Quoted 70 lines · Lorenzo Pieralisi
Quoted 69 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Or have a "validate all parameters" function that you do it in one place?

Quoted 7 lines
Quoted 2 lines

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c
This is bad, thanks Greg for spotting it.

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

The userspace tool (pcitest [1]) that calls these IOCTLS prints "NOT OKAY" on 0 and "OKAY" for 1.

Quoted 8 lines
Quoted 2 lines

I'll change to negative value for error and still keep '0' for data mismatch error?

No, 0 as a return value from a system call means "success". You can not redefine it to mean something else for a tiny single kernel driver, otherwise so much for consistency.

0 is success, -ERROR_NUMBER is an error, please read the ioctl man page.

Thanks Greg!

Mani, Would you be able to post a patch fixing it? If not I'll post it mid next week.

Regards,
Kishon

#6 lore raw
Quoted 95 lines · Kishon Vijay Abraham I

+Mani

Quoted 92 lines · Greg Kroah-Hartman
Quoted 78 lines · Kishon Vijay Abraham I

Hi Lorenzo,

Quoted 75 lines · Lorenzo Pieralisi
Quoted 66 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Quoted 5 lines

Or have a "validate all parameters" function that you do it in one place?

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c
This is bad, thanks Greg for spotting it.

Quoted 8 lines
Quoted 2 lines

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

The userspace tool (pcitest [1]) that calls these IOCTLS prints "NOT OKAY" on 0 and "OKAY" for 1.

I'll change to negative value for error and still keep '0' for data mismatch error?

No, 0 as a return value from a system call means "success". You can not redefine it to mean something else for a tiny single kernel driver, otherwise so much for consistency.

Quoted 7 lines
Quoted 1 line

0 is success, -ERROR_NUMBER is an error, please read the ioctl man page.

Thanks Greg!

Mani, Would you be able to post a patch fixing it? If not I'll post it mid next week.

Will do!

Thanks,
Mani

Quoted 2 lines

Regards,
Kishon

Signature · 1 line
மணிவண்ணன் சதாசிவம்
#7 lore raw
Quoted 98 lines · Manivannan Sadhasivam
Quoted 97 lines · Kishon Vijay Abraham I

+Mani

Quoted 93 lines · Greg Kroah-Hartman
Quoted 78 lines · Kishon Vijay Abraham I

Hi Lorenzo,

Quoted 75 lines · Lorenzo Pieralisi
Quoted 66 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
 	}
 
 	size = param.size;
-	if (size > SIZE_MAX - alignment)
+	if (size > SIZE_MAX - alignment || !size)
 		goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Quoted 5 lines

Or have a "validate all parameters" function that you do it in one place?

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c
This is bad, thanks Greg for spotting it.

Quoted 8 lines
Quoted 2 lines

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

The userspace tool (pcitest [1]) that calls these IOCTLS prints "NOT OKAY" on 0 and "OKAY" for 1.

I'll change to negative value for error and still keep '0' for data mismatch error?

No, 0 as a return value from a system call means "success". You can not redefine it to mean something else for a tiny single kernel driver, otherwise so much for consistency.

0 is success, -ERROR_NUMBER is an error, please read the ioctl man page.

Quoted 8 lines
Quoted 5 lines

Thanks Greg!

Mani, Would you be able to post a patch fixing it? If not I'll post it mid next week.

Will do!

Done! But I haven't fixed the size issue that _this_ patch fixes. I'll leave it up to Shunsuke.

Thanks,
Mani

Quoted 8 lines

Thanks,
Mani

Quoted 2 lines

Regards,
Kishon

Signature · 1 line
மணிவண்ணன் சதாசிவம்
Signature · 1 line
மணிவண்ணன் சதாசிவம்
#8 lore raw

2022年8月19日(金) 23:53 Manivannan Sadhasivam <[email protected]>:

Quoted 106 lines · Manivannan Sadhasivam
Quoted 104 lines · Kishon Vijay Abraham I
Quoted 102 lines

+Mani

Quoted 85 lines · Greg Kroah-Hartman
Quoted 84 lines · Kishon Vijay Abraham I

Hi Lorenzo,

Quoted 70 lines · Lorenzo Pieralisi
Quoted 69 lines · Greg Kroah-Hartman
Quoted 55 lines · Shunsuke Mie

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
      }
 
      size = param.size;
-     if (size > SIZE_MAX - alignment)
+     if (size > SIZE_MAX - alignment || !size)
              goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Or have a "validate all parameters" function that you do it in one place?

Quoted 7 lines
Quoted 2 lines

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/pci/pcitest.c
This is bad, thanks Greg for spotting it.

@Kishon, we have to fix it asap, please let me know if you can post a patch promptly.

The userspace tool (pcitest [1]) that calls these IOCTLS prints "NOT OKAY" on 0 and "OKAY" for 1.

Quoted 8 lines
Quoted 2 lines

I'll change to negative value for error and still keep '0' for data mismatch error?

No, 0 as a return value from a system call means "success". You can not redefine it to mean something else for a tiny single kernel driver, otherwise so much for consistency.

0 is success, -ERROR_NUMBER is an error, please read the ioctl man page.

Thanks Greg!

Mani, Would you be able to post a patch fixing it? If not I'll post it mid next week.

Quoted 5 lines
Quoted 2 lines

Will do!

Done! But I haven't fixed the size issue that _this_ patch fixes. I'll leave it up to Shunsuke.

I'll do that.

Quoted 6 lines

Thanks,
Mani

Quoted 3 lines

Thanks,
Mani

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Regards,
Kishon

--
மணிவண்ணன் சதாசிவம்

--
மணிவண்ணன் சதாசிவம்

Thanks,
Shunsuke

#9 lore raw
In reply to Greg Kroah-Hartman

2022年8月16日(火) 19:27 Greg Kroah-Hartman <[email protected]>:

Quoted 67 lines · Shunsuke Mie
Quoted 55 lines

Although dma_map_single() doesn't permit zero length mapping, the each test functions called the function without zero checking.

A panic was reported on arm64:

[   60.137988] ------------[ cut here ]------------
[   60.142630] kernel BUG at kernel/dma/swiotlb.c:624!
[   60.147508] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   60.152992] Modules linked in: dw_hdmi_cec crct10dif_ce simple_bridge rcar_fdp1 vsp1 rcar_vin videobuf2_vmalloc rcar_csi2 v4l
2_mem2mem videobuf2_dma_contig videobuf2_memops pci_endpoint_test videobuf2_v4l2 videobuf2_common rcar_fcp v4l2_fwnode v4l2_asyn
c videodev mc gpio_bd9571mwv max9611 pwm_rcar ccree at24 authenc libdes phy_rcar_gen3_usb3 usb_dmac display_connector pwm_bl
[   60.186252] CPU: 0 PID: 508 Comm: pcitest Not tainted 6.0.0-rc1rpci-dev+ #237
[   60.193387] Hardware name: Renesas Salvator-X 2nd version board based on r8a77951 (DT)
[   60.201302] pstate: 00000005 (nzcv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[   60.208263] pc : swiotlb_tbl_map_single+0x2c0/0x590
[   60.213149] lr : swiotlb_map+0x88/0x1f0
[   60.216982] sp : ffff80000a883bc0
[   60.220292] x29: ffff80000a883bc0 x28: 0000000000000000 x27: 0000000000000000
[   60.227430] x26: 0000000000000000 x25: ffff0004c0da20d0 x24: ffff80000a1f77c0
[   60.234567] x23: 0000000000000002 x22: 0001000040000010 x21: 000000007a000000
[   60.241703] x20: 0000000000200000 x19: 0000000000000000 x18: 0000000000000000
[   60.248840] x17: 0000000000000000 x16: 0000000000000000 x15: ffff0006ff7b9180
[   60.255977] x14: ffff0006ff7b9180 x13: 0000000000000000 x12: 0000000000000000
[   60.263113] x11: 0000000000000000 x10: 0000000000000000 x9 : 0000000000000000
[   60.270249] x8 : 0001000000000010 x7 : ffff0004c6754b20 x6 : 0000000000000000
[   60.277385] x5 : ffff0004c0da2090 x4 : 0000000000000000 x3 : 0000000000000001
[   60.284521] x2 : 0000000040000000 x1 : 0000000000000000 x0 : 0000000040000010
[   60.291658] Call trace:
[   60.294100]  swiotlb_tbl_map_single+0x2c0/0x590
[   60.298629]  swiotlb_map+0x88/0x1f0
[   60.302115]  dma_map_page_attrs+0x188/0x230
[   60.306299]  pci_endpoint_test_ioctl+0x5e4/0xd90 [pci_endpoint_test]
[   60.312660]  __arm64_sys_ioctl+0xa8/0xf0
[   60.316583]  invoke_syscall+0x44/0x108
[   60.320334]  el0_svc_common.constprop.0+0xcc/0xf0
[   60.325038]  do_el0_svc+0x2c/0xb8
[   60.328351]  el0_svc+0x2c/0x88
[   60.331406]  el0t_64_sync_handler+0xb8/0xc0
[   60.335587]  el0t_64_sync+0x18c/0x190
[   60.339251] Code: 52800013 d2e00414 35fff45c d503201f (d4210000)
[   60.345344] ---[ end trace 0000000000000000 ]---

To fix it, this patch adds checkings the payload length if it is zero.

Fixes 343dc693f7b7 ("misc: pci_endpoint_test: Prevent some integer overflows")
---
 drivers/misc/pci_endpoint_test.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/misc/pci_endpoint_test.c +0 -0
index 8f786a225dcf..d45426a73396 100644

drivers/misc/pci_endpoint_test.c static bool pci_endpoint_test_copy(struct pci_endpoint_test *test, · line 368

Quoted 6 lines
drivers/misc/pci_endpoint_test.c +1 -1
@@ -364,7 +364,7 @@ static bool pci_endpoint_test_copy(struct pci_endpoint_test *test,
      }
 
      size = param.size;
-     if (size > SIZE_MAX - alignment)
+     if (size > SIZE_MAX - alignment || !size)
              goto err;

Can we test size first? And do it in the ioctl handler in one place so you don't have to add it everywhere?

Quoted 2 lines

Or have a "validate all parameters" function that you do it in one place?

Yes, it is reasonable I think. I'll unify validations to the function.

Quoted 5 lines

Also, all of these ioctl handlers are wrong, they are returning "0" to userspace if they fail, which is not correct at all. thanks,

greg k-h