Re: [PATCH 1/2] HID: core: i2c-hid: fix size check and type usage
flat view
From: Benjamin Tissoires <hidden>
Date: 2018-01-04 10:44:07
Also in:
lkml
Hi Aaron, There are quite some changes I'd like to see in this patch. See below. On Tue, Jan 2, 2018 at 6:30 PM, Aaron Ma [off-list ref] wrote:
When convert char array with signed int, if the inbuf[x] is negative then upper bits will be set to 1. Fix this by using u8 instead of char. ret_size has to be at least 3, hid_input_report use it after minus 2 bytes. size should be more than 0 to keep memset safe. Cc: stable@vger.kernel.org Signed-off-by: Aaron Ma <redacted> --- drivers/hid/hid-core.c | 4 ++-- drivers/hid/i2c-hid/i2c-hid.c | 10 +++++-----
I'd like to have this patch at least split in 2. One for hid-core, one for i2c-hid. Both files are not targeting the same issue, so it would make sense to not have them in the same patch.
quoted hunk ↗ jump to hunk
2 files changed, 7 insertions(+), 7 deletions(-)diff --git a/drivers/hid/hid-core.c b/drivers/hid/hid-core.c index 0c3f608131cf..992547771d96 100644 --- a/drivers/hid/hid-core.c +++ b/drivers/hid/hid-core.c@@ -1506,7 +1506,7 @@ int hid_report_raw_event(struct hid_device *hid, int type, u8 *data, int size, if (rsize > HID_MAX_BUFFER_SIZE) rsize = HID_MAX_BUFFER_SIZE; - if (csize < rsize) { + if ((csize < rsize) && (csize > 0)) {
I would think a call to this function with a csize <= 0 is a reason to fail. And actually, I think an other thing we shouold do is changing the prototype to have an unsigned int for size instead of a simple int. Tracking the impact of such change might involve a bigger patch than a simple check here, but we should be able to at least have the compiler complaining if some driver starts using negative values.
quoted hunk ↗ jump to hunk
dbg_hid("report %d is too short, (%d < %d)\n", report->id, csize, rsize); memset(cdata + csize, 0, rsize - csize);@@ -1566,7 +1566,7 @@ int hid_input_report(struct hid_device *hid, int type, u8 *data, int size, int i report_enum = hid->report_enum + type; hdrv = hid->driver; - if (!size) { + if (size <= 0) {
This should also be solved by changing the prototype of the function.
quoted hunk ↗ jump to hunk
dbg_hid("empty report\n"); ret = -1; goto unlock;diff --git a/drivers/hid/i2c-hid/i2c-hid.c b/drivers/hid/i2c-hid/i2c-hid.c index e054ee43c1e2..09404ffdb08b 100644 --- a/drivers/hid/i2c-hid/i2c-hid.c +++ b/drivers/hid/i2c-hid/i2c-hid.c@@ -144,10 +144,10 @@ struct i2c_hid { * register of the HID * descriptor. */ unsigned int bufsize; /* i2c buffer size */ - char *inbuf; /* Input buffer */ - char *rawbuf; /* Raw Input buffer */ - char *cmdbuf; /* Command buffer */ - char *argsbuf; /* Command arguments buffer */ + u8 *inbuf; /* Input buffer */ + u8 *rawbuf; /* Raw Input buffer */ + u8 *cmdbuf; /* Command buffer */ + u8 *argsbuf; /* Command arguments buffer */ unsigned long flags; /* device flags */ unsigned long quirks; /* Various quirks */@@ -473,7 +473,7 @@ static void i2c_hid_get_input(struct i2c_hid *ihid) ret_size = ihid->inbuf[0] | ihid->inbuf[1] << 8; - if (!ret_size) { + if (ret_size <= 2) {
Please do a separate case, after this one. RESET is acked by a size of 0, a size of 1 or 2 is a bug that needs to be fixed from the HW point of view. Cheers, Benjamin
/* host or device initiated RESET completed */
if (test_and_clear_bit(I2C_HID_RESET_PENDING, &ihid->flags))
wake_up(&ihid->wait);
--
2.14.3